2018 Thesis Dana Doherty - Chatbots

L
EEE521 Final Year Project Report

School of Computing, Engineering &
Intelligent Systems
Dana Doherty
BSc Hons Computer Science
Task-based Interaction Chatbot
Supervisor Dr. Kevin Curran

Second Marker Dr. Daniel Kelly
Plagiarism Statement
I declare that this is all my own work and does not contain unreferenced material copied from any
other source. I have read the University’s policy on plagiarism and understand the definition of
plagiarism. If it is shown that material has been plagiarised, or I have otherwise attempted to obtain an
unfair advantage for myself or others, I understand that I may face sanctions in accordance with the
policies and procedures of the University. A mark of zero may be awarded and the reason for that
mark will be recorded on my file.
I confirm that the Originality Score provided by TurnItIn for this report is ________.
[your signature]
Acknowledgements
I would like to take this opportunity to thank my project supervisor Dr Kevin Curran for his
continuous guidance and encouragement throughout the entirety of the project. I would also like to
thank Dr Daniel Kelly for his feedback and guidance during the development stage. Finally I would to
thanks Dr Michael McTear for his guidance and valuable and critical input on the Testing and
Evaluation Chapters.
Contents
1. INTRODUCTION ................................................................................................................. 1
1.1 AIMS & OBJECTIVE ......................................................................................................... 1
1.2 LITERATURE REVIEW .................................................................................................... 2
1.3 CHATBOTS IN INDUSTRY .............................................................................................. 3
1.3.3 CHATBOTS ..................................................................................................................... 6
1.4 NATURAL LANGUAGE UNDERSTANDING ENGINE ................................................ 7
1.5 ARTIFICIAL INTELLIGENCE .......................................................................................... 8
1.5.1 ARTIFICIAL INTELLIGENCE METHODS .................................................................. 8
1.6 CHATBOT ARCHITECTURE ......................................................................................... 10
1.7 SUMMARY OF REPORT ................................................................................................ 10
1.8 CONCLUSION .................................................................................................................. 11
2. REQUIREMENTS SPECIFICATION AND ANALYSIS .................................................. 11
2.1 PROBLEM STATEMENT ................................................................................................ 11
2.2 PROPOSED SOLUTION .................................................................................................. 12
2.3 SOFTWARE DEVELOPMENT METHODOLOGIES .................................................... 12
2.3.1 WATERFALL METHODOLOGY ................................................................................ 13
2.3.2 INCREMENTAL MODEL............................................................................................. 13
2.3.3 CHOSEN METHODOLOGY ........................................................................................ 14
2.4 INFORMATION GATHERING TECHNIQUES ............................................................. 14
2.4.1 CHOSEN INFORMATION GATHERING TECHNIQUE ........................................... 16
2.5 FUNCTIONAL AND NON-FUNCTIONAL REQUIREMENTS .................................... 17
2.5.1 FUNCTIONAL REQUIREMENTS ............................................................................... 18
2.5.2 NON-FUNCTIONAL REQUIREMENTS ..................................................................... 18
2.6 SOFTWARE AND HARDWARE SPECIFICATION ...................................................... 19
2.6.1 SOFTWARE REQUIREMENTS ................................................................................... 19
2.6.2 HARDWARE REQUIREMENTS.................................................................................. 19
2.7 PROJECT PLANNING ..................................................................................................... 19
2.7.1 PROJECT MILESTONES AND DELIVERABLES ..................................................... 20
2.7.2 DEVELOPMENT ITERATION BREAKDOWN .......................................................... 21
2.7.3 RISK MANAGEMENT.................................................................................................. 22
2.7.4 PROJECT MANAGEMENT .......................................................................................... 23
3. DESIGN ............................................................................................................................... 25
3.1 ARCHITECTURAL DESIGN........................................................................................... 25
3.2 DATABASE DESIGN....................................................................................................... 29
3.3 CLASS DIAGRAM ........................................................................................................... 30
3.4 SPEECH RECOGNITION ................................................................................................ 32
3.5 ACTIVITY DIAGRAMS .................................................................................................. 33
3.6 SEQUENCE DIAGRAMS ................................................................................................ 38
3.7 CONVERSATIONAL USER INTERFACE DESIGN ..................................................... 40
3.8 DIALOG DESIGN............................................................................................................. 42
4. IMPLEMENTATION AND TESTING .............................................................................. 46
4.1 CONFIGURING THE DEVELOPMENT ENVIRONMENT: ......................................... 46
4.2 AUTHENTICATION: ....................................................................................................... 48
4.3 DATABASE DEVELOPMENT ........................................................................................ 48
4.4 GOOGLE TWO-FACTOR AUTHENTICATION: .......................................................... 49
4.5 OPEN BANKING API: ..................................................................................................... 51
4.6 WEBHOOK DEVELOPMENT: ....................................................................................... 56
4.7 CHATBOT WEB CLIENT................................................................................................ 63
4.8 GIT VERSION CONTROL............................................................................................... 64
5. TESTING ............................................................................................................................. 64
5.1 SIMULATED CONVERSATIONAL TESTING ............................................................. 65
5. 2 WEBHOOK TESTING .................................................................................................... 73
5.3 USER TESTING:............................................................................................................... 74
6. EVALUATION................................................................................................................... 75
6. 1 CONCLUSION ................................................................................................................. 79
6.2 FUTURE ADVANCEMENTS ......................................................................................... 79
6. REFERENCES ...................................................................................................................... 2
HTTPS://WWW.BBA.ORG.UK/WP-
CONTENT/UPLOADS/2014/06/BBA_COMPETITION_REPORT_23.06_WEB_2.0.PDF .. 4
7.APPENDICES ........................................................................................................................ 8
7.1 APPENDIX A QUESTIONNAIRE RESULTS USER REQUIREMENTS .................... 9
EEE521 Final Year Project 2017/18 B00659303
Abstract
People interact with systems more and more through voice assistants and chatbots. The days of solely
engaging with a service through a keyboard are over. These new modes of user interaction are aided in
part by This research will investigate how aadvancements in Artificial Intelligence and
Machine Learning technology.
are being used to improve many services. In

particular it will look at the development of chatbots as a channel for information
distribution.
This project aimed to implement a web-based chatbot to assist with online

banking, using tools that expose artificial intelligence methods such as natural language understanding.
Allowing users to interact with the chatbot using natural language input and to train the chatbot using
appropriate methods so
it will be able to generate a response. The chatbot will allow users to
view all their personal banking information all from within the chatbot.
The produced prototype was found to be a very The TrueLayer API will be used to retrieve account
information which has
just recently been made accessible to the public. This is part of the Open Bank
Standard which allows developer’s access banking data through the API
Rest client. This came about through the bank as a platform (BaaP) standard.
useful tool to justify the need of a modern method of interaction to be integrated within many services
offered by banks and financial businesses. In an industry with low user satisfaction rates and limited
technology to increase accessibility. It is clear the chatbot overcomes the challenges banks face to
increase the use of their services and gain a competitive edge over leading competitors.
With many people adopting Smart Assistant Devices such as Google Home or Amazon’s Alexa. The
chatbot was tested across a range of devices such as Google Home and Assistant on android devices to
outline the key differences between the two modes of interaction , spoken and text dialog. These test
were carried out to identify the value in integrating such technology surrounding the recent interest in
1
chatbots and conversational interfaces. Proving chatbots can be applied to a specific domain to enhance
accessibility, reaffirming that they are more than just a passing fad and have a viable use.
2
1. Introduction
“Digitalisation, the surge of mobile and internet connected devices has revolutionised the way people
interact with one another and communicate with businesses” (Eeuwen, M.V. ( 2017)). Millennials are
accepting and supporting new technology into the routine of their everyday life, this is becoming more
is becoming more prevalent as technology companies are streamlining Artificial

IntelliIntelligence (AI) into the products they offer, such as; Google Assistant, Google
Home and Amazon Alexa. The new and upcoming generation are expected to be
critical and game changing customers for businesses. “They demand effortless experiences, answers
within seconds, not minutes and more intelligent self-service options” (Teller Vision,, . ( 2017)).
The banking and the financial service industry was one of the first industries to adopt technology. This
integration has grown massively, helping banks reach a wider customer base enabling them to perform
their banking conveniently (Baptista and , G. and Oliveira, , T. (2015)).
Banks are becoming ever more competitive with each other to adopt the newest advancements in
technology to provide an improved delivery service to satisfy customers. Ulster Bank, Deloitte, AIB
and PTSB are wanting to focus on integrating new technology to improve the speed at which
transactions are acknowledged (Global Banking News, . (, 2017)). With this in mind the relationship
with the customer is always evolving due to the growth of technology.
Banks are now enabling the use of technology so customers can perform more tasks online, such as;
cheque image clearing to allow the payment of cheques remotely and intelligent chatbots to increase
customer service and assist employees. A chatbot is a “simple software program that can respond to
customer prompts i.e. what’s my bank balance?” (Entrepreneur, . (, 2016)). Mastercard has launched
Kai an artificial intelligent chatbot and other bots for financial services. They can handle customer
queries such as: ‘what is APR?’, requests, look at spending habits and solve problems. This in turn
enables financial institutions to provide a new, engaging experience and strengthen their relationship
with the customer, with the aid of natural language used by bots to establish a more personal and
contextual conversation (Wire, . (, 2016)). The focus of this project is to implement these new
technologies to create an intelligent chatbot to enable banks to appeal to millennials and potentially gain
a lifelong customer.
1.1 Aims & Objective

This work aims to provide a fast and convenient way to manage your banking. The online banking
chatbot will help facilitate the user with queries and assist with personal banking.
1
The application will allow users to c:

Check their bBalance,
View aAccount dDetails & s
View Ttransactions. There will be integration of SMS, currency conversions, two factor authentication
and the ability to interact with the service through a chatbot.
• SMS confirmations and email mini statement schedule. The Starling

API will be used to access the users banking information requested
through the chatbot.
• Currency conversions
• Ask queries through chatbot and get appropriate and immediate
responses
• Improve customer service through conversation using the chatbot
The chatbot will be implemented using the Laravel Framework and Dialogflow.
The Dialogflow will be utilised as the NLU to perform artificial intelligence
methods such as; Natural Language Processing (NLP), POS tagging, and entity
recognition to analyse text and carry out the appropriate actions. Dialogflow is a
Natural Language Understanding Engine (NLU) used for the extraction of entities
and intent from a user’s message.
Google Two-Factor Authentication will be implemented as an extra measurement
of security for the customer. This ensures that no one else can access their account
and view their personal banking information. A unique code will be generated
and sent to their phone from scanning a QR code displayed after a successful
login. Users will download the Google Authenticator app onto their mobile
device to receive the unique code generated which will grant access to the chatbot.
1.2 Literature Review

The banking industry has multiple electronic delivery channels in use to distribute technology assets
and services for the benefit of their customers. Online banking is a commodity of commerce within
2
financial services as well as banking industries (Ajimon and , G. G.S. Gireesh K,.(George and Kumar,
2013)). Advancements in technology has transformed many of our services into the digital era and the
banking industry is one of the primary industries to avail of these advancements to improve their
services. Currently within the UK two paradigms are available for online banking. One of which is an
integrated internet bank which still operates through the branch but has an online presence. The other,
a stand-alone internet bank, that operates completely independently and its only existence is solely
through the internet (MarketLine, 2017).
Banks implement technology to strengthen their processing capacity, acquire a larger customer base
and expand the services they could offer (Consoli, . (, 2005)). The use of internet banking has grown in
demand enormously in the last decade. “15% of branch customers use online banking once a day, 59%
once a week, 77% at least once a month and 53% were confident in carrying out the best part of their
banking online” (Barty, J. and Recketts, T. BBA, (2014)). Online banking has become more popular as
it negates the need for customers to visit their local branch as they can manage their finances on the go
to meet the demand of modern life. This is evident as branchless banks are now emerging from the
industry such as Atom Bank and many banks now closing some of their branches. This is evident with
the recent closure of 11 Ulster Bank Branches in NI due to the increased number of customers
performing their banking online (The Belfast Telegraph).
HoweverHowever, a recent study by Ling et al., (2016), notes that most internet banking service
providers struggle to get many of their customers to use their service. They identify lack of customer
satisfaction when using online banking services to be a major cause. “Service quality, web design and
content, security, privacy, speed and convenience” (Ling et al., 2016) are stated identified as the top
factors influencing customer satisfaction This suggests that there is a lack of technology in place to
enhance the customer online banking experience which could be improved by integrating a chatbot to
provide an efficient, convenient and personal service.
1.3 Chatbots in Industry

Most businesses and organisations are understanding the potential benefits of machine learning and
artificial intelligence to have a positive change on how they perform business. Artificial intelligence
has progressed to allow the development of more sophisticated chatbots. Organisations are focusing on
specific areas of user engagement that take up a lot of time but can be replaced through the use of a
chatbot. Chatbots can understand what the customer needs from a single text instead of the customer
having to follow a process of multiple steps.
Chatbots are used to automate customer service and reduce manual tedious tasks performed by
employees so they can spend their time more productively on higher priority tasks. Establishments that
regularly deal with its customers have discovered the potential of chatbots as a channel to distribute
3
more efficient and immediate information to customers in comparison to a customer service

representative regarding queries and issues (Onufreiv, YY. , 2017).
HDFC Bank has merged with Niki.ai an artificial intelligence company to develop a state of the art
conversational banking chatbot ((HDFC Bank,. , 20177)). The chatbot is accessible through the banks
Facebook Messenger allowing users to utilise e-commerce and banking transactions all within the
chatbot.
There is also a chatbot integrated within the HDFC login page to assist in online banking. The chatbot
was developed as a concierge service, this is definitely one approach that can increase customer
satisfaction within online banking enabling banks to develop a better relationship with its customers.
Chatbots will renew and modernise the customer service industry and the main sectors outlined
including the banking and healthcare sector (Newswire, ., 2017). Industries such as; retail, healthcare,
e-commerce and banking, are expected to achieve considerable savings from integrating a chatbot.
Between 2017 and 2020 chatbots are estimated to make savings of $8 billion per annum for businesses.
The majority of savings will be made through customer service as customers can now ask queries about
banking through the chatbot rather than having to call the bank, allowing banks to save on call centre
staff. The study predicted that the integration of chatbots within the banking industry would rise from
12% to 90% by 2022.
Most online banking services would benefit from having a chatbot integrated into their services. The
use of bots help internet banking service providers establish a better relationship with its customers.
Customers can get answers to query’s immediately, conduct e-commerce and banking operations all
from within the one bot conversation. This is another benefit of using an online banking system with
and integrated chatbot as “87% customers that bank online prefer to execute their personal finance
operations within a single site” (Dauda, et al., 2015).
W.B. King, W.B. (2017), records a survey conducted by Personetics which identified “40% of its
clients are planning to integrate a bot into its services within the next 2 to 3 years and 70% said AI was
an exciting opportunity”. This clearly identifies the demand and need for a chatbot within the financial
services Industry.
4
Figure 1: Figure 1.3.1 - Interest in chatbots within the Financial Industry (Source: W.B. King, W.B
2017)
Mark ZuckerBerg CEO of Facebook, believesC chatbots provide more personal and immediate ways
for customers to communicate with an organisation allowing them to provide consistent and appropriate
customer service at scale (Conversational Business, 2017). Facebook has now released its own
Messenger Bots, which allows organisations to develop their own Chatbot through Messenger to deliver
their services.
Facebook are also developing AI methods to obtain specific information related to users across relevant
topics. The acquired information is used as a dataset to develop and improve their Artificial Intelligence
so they can analyse, group and rate each user post (Rossmann et al., 2017).
“44% of US consumers prefer to interact with a chatbot compared to a human in relation to customer
queries” (Wellers et al., 2017). This is evident as Chatbots are also becoming more prevalent in other
industries such as; insurance, recruitment, media and pharmaceuticals.
5
Figure 1.3.2: Industries piloting chatbots (Source: Etlinger,. , 2017)
Figure 1.3.2 2. displays the tasks that can be replaced with the integration of chatbots into an
organisation, with the goal of advancing access and use over time.
1.3.3 Chatbots
A chatbot is a software tool that utilises natural language processing (NLP) for human machine
interaction (HMI) and Machine Learning (ML). “The complexity of a chatbot is directionally
proportional to the scope of the domain”. An open domain requires a larger knowledge base, whereas,
a closed domain has a more specific knowledge base that was developed to achieve a specific goal
(Gregori, E. , 2017).
6
Chatbot technology initially began in the 1960s to determine whether a chatbot could be portrayed as a
human. Throughout the 1980s there was an elevated amount research carried out on natural language
interfaces which lead to the development of sophisticated chatbot architectures such as A.L.I.C.E. This
chatbot architecture is one of the earlier chatbots developed in 1995 by Dr Wallace which is now open-
source, the acronym stands for Artificial Linguistic Internet Computer Entity. This is a chatbot you can
create through interaction as it will learn from previous interactions to create its knowledge base. Its
knowledge is saved in AIML (Artificial Intelligent Mark-up Language) files which evolved from the
Extensible Mark-up Language (XML) (Shawar, B.A and and Atwell, E, 2007).
Chatbots are developed using two approaches; a rule based approach where chatbots operate by moving
through branches of a tree diagram of an expert system. The second approach involves advanced
artificial intelligence and machine learning, thus the chatbot can learn from the conversations,
generating appropriate responses to continuously improve over time (Watson, A. 2017).
There are two modes in which chatbots can simulate a conversation with users which include :
Ssystem-initiated chatbots where– they commence the conversation with the user and
User-initiated chatbots where- the user directs the conversation instead.
Systems that incorporate the two methods of initiation are known as mixed initiative systems (Duijst,
D. 2017).
1.4 Natural Language Understanding Engine

The chatbot engine is thought of as one of the most critical elements of a chatbot, alias “Natural
Language Understanding (NLU) engine” ( Kar, R and Haldar, R. 2016). The NLU holds liability for
the translation of conversational dialogs to actions which are understood by the machine. NLU engines
use a variety of artificial intelligence methods to understand the natural language used in conversational
interfaces such as chatbots. These methods consist of: Natural Language Processing (NLP) and Machine
Learning (ML) (Kar, R and Haldar, R. 2016).
Googles Dialogflow, previously known API.ai, is a natural language understanding engine that
identifies the intent and context from the natural language in user supplied utterances. These concepts
are used to develop the behaviour of the chatbot and how coherently it interacts with the user. Intents
are used to establish a connection between the user input and the appropriate action to be executed by
the chatbot in order for the user to achieve their goal. Contexts are utilised to differentiate and
understand user input which may have an alternative meaning depending upon the current
conversational context (Gregori, E. 2017).
7
1.5 Artificial Intelligence

“Artificial Intelligence is neither a new technology nor a machine”. Artificial intelligence is the
recognition of outcome-direction which is the rapid analysis of live data to achieve the expected goal.
Outcome-directed thinking splits from the confines of the rule-directed approach that is accomplished
through artificial intelligence.
The generalised practice of AI can be broken down into a straightforward process which does not
require an experienced level of proficiency to understand. First of all, a numerical representation is
established for the target or outcome. Specific data is then associated with the target is gathered and
conditions and behaviours are investigated to increase the likelihood of achieving the expected target.
Multiple aspects can determine the outcome. The weight of each aspects effect is computed. “AI uses
the relative weighting of each aspect to create a prediction (evaluation) formula” (Yano, K. 2017).
Lastly, the formula devised from the weighted aspects are employed to business decisions (Yano, K.
2017). AI can be classified into four groups: “systems that think like humans, systems that act like
humans, systems that think rationally and systems that act rationally” ((Russell, S.J and Norvig, P.
1995).
AI is generally categorised as strong and weak AI: strong AI is the production of human-like intelligent
systems. Weak AI would be the integration of intelligent algorithms embedded within a system.
“Machine learning, deep-learning, natural
language processing and neural networks are often summarised under the term of AI” (Exner-Stöhr, et
al., 2017)
1.5.1 Artificial Intelligence Methods

“Natural Language Processing (NLP) is a theory motivated range of computational techniques, for the
automatic analysis and representation of human language” (Jurafsky , D and Martin, J.H. 2017). Natural
Language Processing technology has made great advancements in machine learning based systems to
be able to extract meaning from natural language utterances also known as sentiment analysis (Cambria
and White, 2014).
There are many techniques used in NLP for the analysis of natural language include Entity recognition:
which Entity Recognition is a technique which rrecognises entities in text. The most common entities
include; nouns, organisations, people and places.
Named Entity Recognition (NER) is the task of finding every instance of a named entity in text and
label its type in order to classify it correctly (Jurafsky , D and Martin, J.H. 2017).
8
“Entities are domain specific information extracted from the utterance that maps the natural language
phrases to their canonical phrases in order to understand the intent. They help in identifying the
parameters which are required to take specific action” (Kar, R and Haldar, R. Kar and Haldar, 2016).
Establishing the context of the of the users message is a vital feature that allows the chatbot to deal with
situations that it may not be able to carry out a specific action for. This is due to the user input being
very vague or may have an alternative meaning. The context is the capacity of a chatbot to sustain its
state, also referred to as the number of user supplied input (utterances) when the context is extracted
and the appropriate intent is paired to conduct the desired action for the user.
Intents are the core backbone of conversational interfaces. The intents represent what the customer is
wanting to achieve such as ‘show me my balance’. The text sent by the user in natural language is
analysed to identify the corresponding intent of the text. This requires matching a specific user supplied
phrase with an appropriate action to be executed by the system. The chat bot would then return
appropriate dialog in order for the user to achieve their goals.
Actions are the processes or steps executed by the system when the intent of a message is identified,
they contain parameters which categorise and define its properties (Kar, R and Haldar, R. Kar and
Haldar, 2016).
Sentiment analysis incorporates multiple natural language processing techniques in order to extract
meaning and polarity from text. Polarity detection classifies text to be either positive or negative and
measures the intensity of the overall polarity detected. Sentiment analysis achieves a more in depth
understanding of the contextual role of each concept within a given piece of text (Cambria, E. and
White, B, 2017).
Part of Speech (POS) Tagging: this involves assigning a label to each word in the user
input with its part of speech (e.g. noun, verb, adjective, etc.). The labels or tags can be
rule-based (a manually developed set of rules is defined to establish part of speech
for ambiguous words provided in the conversational context). The labels can also be
developed utilising advanced models that are trained using input labelled with the
appropriate POS. This is additionally used in response generation in order to outline
the POS object type of the expected response made by the chat bot (Cahn, 2017).
9
1.6 Chatbot Architecture

Previously chatbots solely supported a single adjacency pair, also known as a one-shot conversation.
However, modern chatbots can sustain multiple adjacency pairs, remembering states and contexts
between conversations and have the capability to associate data in different adjacency pairs which is
related. This is the bots ability to preserve the conversation. A chatbot consists of four main parts: front-
end, knowledge-base, back-end and corpus which is the training data. The front end is accountable for
enabling communication between the bot and the user. The NLU utilises Artificial intelligence methods
to identify the intent and context of the user input. An appropriate response is generated from the users’
intent. The knowledge base defines the chatbots knowledge, which is created within the NLU and
supported by the back-end, the back-end applies the domains corpus to produce the knowledge base
(Gregori, E. 2017).
Figure 31.6.1: illustrates a standard chatbot architecture (source: Michael McTear, (2018) ).
Input can be supplied to the chatbot in the form of text or speech. The Input is sent to the dialog
management system which is the NLU in this case, which determines an appropriate response and
amends the chatbots state accordingly to carry out the required action. The chatbot will produce text
and speech responses in the form of both text and speech.
1.7 Summary of Report
The remainder of this report will focus on:
• Analysis - This section will include various software methodologies; quantifiable feedback will
be collected from questionnaires which will be analysed to identify the requirements of the
chatbot. The internal and external hardware and system requirements are outlined. A Gantt
chart will be provided outline the project schedule and when each milestone should be met.
10
• Design - This section will cover the overall design of the chatbot; UI diagrams/story boards
will be included which detail how the GUI will appear to the user. The technical architecture
will be designed and displayed in graphical form such as an activity diagram. It will outline
how each component interacts with each other. The design for the database and the relationships
between tables will be provided.
• Implementation and testing - The section will document the implementation process to develop
the solution, including; the code used to develop compelling features. Define what was learnt
during each iteration and document each prototype. Once implementation is complete the
chatbot will be tested using unit testing and appropriate test cases. Any errors or faults identified
will be fixed to ensure maximum quality.
• Evaluation and Reflection - An evaluation will be carried on the completed project to determine
its quality and value. It will detail how the overall project was developed. Reflecting upon the
overall experience, highlighting any areas that could be improved for future work and what
well throughout the project.
1.8 Conclusion
There is clear evidence based on the research conducted that there will be a drastic increase in the
number of chatbots being implemented within the financial service industry. The vast amount of
research that has been carried out, and currently ongoing, within the artificial intelligence field has led
to the rise of more sophisticated and intellectual chatbots. This will prove to be immensely beneficial
in providing convenient and accessible customer service at a rapid scale.
2. Requirements Specification and Analysis

This section provides a clear analysis for the potential solution, to outline and breakdown the
necessary requirements to implement the chatbot. This includes:
• the problem the project is overcoming

• the design methodologies considered for implementation
• functional and non-functional requirements
• software and hardware utilised to implement the project
2.1 Problem Statement

It is evident from the research carried out in the literature review that modern financial
11
services are constantly seeking to expand their technologies, both to improve customer service and
increase delivery of services through the advancements in technology. This is to gain a competitive
edge over other banks for financial benefits and to expand its customer base. A domain specific chatbot
will be implemented to assist users with their banking. In order to overcome the user satisfaction issues
associated with online banking services. The chatbot will provide personal and efficient communication
between the user and their bank in order to manage their finances and get assistance when needed, such
as; answering any queries and booking appointments. The chatbot will allow users to feel confident
and comfortable when using this service regardless of the user’s computer literacy due to the natural
language used in messages. It also provides a very accessible and efficient service as all interactions
will take place within the one chat conversation negating the need for the user to navigate through a
site.
2.2 Proposed solution

The proposed solution is to create a chatbot to simulate a human conversation
to assist users with their banking needs and to provide a more personal experience. Advancements in
artificial Intelligence, machine learning techniques, improved aptitude for decision making, larger
availability of domains and corpus, have increased the practicality of integrating a chat bot into
applications (Dole et al., 2015).
Users will be able to ask any banking related queries in natural language that they are comfortable
using such as; view account information, transactions and check balance. The chatbot will identify and
understand what the user is asking and generate an appropriate response based on the conversational
context. Immediate responses will be provided by the chatbot to redeem the need for the user to have
to call or visit their local banks branch and wait in queue in order to get through to an advisor for
assistance.
In order to make the application more secure Googles 2 Factor Authentication will be integrated to
increase security ensuring only registered users can gain access to their account preventing the risk of
fraud.
2.3 Software Development Methodologies

Deciding upon an appropriate methodology is vital for the overall development of any software
application to ensure a realistic timeframe is established for each stage of the project and requirements
are clearly outlined. Various development methodologies will be discussed and considered for the
development and design of this software. This section will highlight the development methodology that
is best suited to this project.
12
2.3.1 Waterfall Methodology

This is a very traditionally methodology, which is usually introduced when you initially
learn about software development. The waterfall model is a very predictive approach
to software development that consists of 5 stages to include; requirements gathering,
analysis, design, implementation and testing. Each stage is completed subsequently

of one another. A major drawback of the waterfall model is that it is very inflexible, as
the project is broken up into phases. Each phase is given a deadline in order for a
deliverable to be produced at the end of each phase to adhere to the overall project schedule. The
success and progression of the project is measured from the project deliverables, design documents and
test plans. As each phase of the project is outlined at the beginning of the project lifecycle and targets
have been set it’s difficult to integrate new requirements or a change in requirements that may be
identified at a later stage as it would adversely affect the overall project schedule. The waterfall model
moves a lot of the more high risk and difficult components towards the end of the project life cycle (Nat
Laundry, N. 2011).
2.3.2 Incremental Model

This software methodology evolved from the waterfall model. The application is designed, developed
and tested using iterative incremental build stages. At the end of each build a subsystem or feature will
be created. The project will progress in complexity as new requirements are likely to be discovered and
implemented in each
incremental build, developing on top of the functionality from the last build leading to the overall
development of the application.
It is very common for software to be released in stages, it is critical that component

versions utilised within the software are managed throughout the entire lifecycle using
version control tools such as GitHub. Each build will only last a few weeks to produce
a baseline version of the application. Feedback can be given on any requirement
errors or faults found in the application. Distributing the development of the project
over various build cycles can lower the risks associated with development to a more
manageable level as requirements are broken down into smaller functionality to be
implemented at the end of each build.
13
Figure 42.3.2.1: . Displaying the Incremental Model (Source: Moniruzzaman and Hossain , 2013)
Testing and debugging are made easier as faults are identified early on during small
manageable cycles. This methodology is flexible during implementation as new
requirements identified are easily integrated at each build and a, updated version is
released (Khan et at., 2011).
2.3.3 Chosen Methodology

The incremental model is the most suitable development methodology to implement
for this project. The flexibility of the incremental model makes it ideal for this project
as it is likely new requirements will be identified during the later stages of development
and each iterative build makes it easy to implement new requirements throughout the
the development process.
2.4 Information Gathering Techniques

To determine the requirements of an application, information gathering must be carried
out in order to know what the end user wants the system to be able to do and what
the user hopes to achieve through using the application. There are many information
gathering techniques, these include :
Questionnaires/Surveys: this method of information gathering is efficient in collecting
14
data of a large group quickly. Surveys consist of

a range of questions such as closed ended questions i.e. where the user is prompted
to select a choices/boxes. Open-ended questions allow user to express their opinion
or provide whatever feedback they feel is suitable. Other types of questions can be
used such as ranges and scales which can be used to gauge how strongly a user may feel. Surveys can
be disturbed through email, online, paper-based and telephone surveys. The questions used must be
clear and precise. Surveys should be tested on a small focus group of user to get feedback on the survey
this ensures the questions are unambiguous and unbiased (Jonathan Lazar, J. 2001).
Interviews : interviews are a structured approach to information gathering. They are

conducted face to face to allow the interviewer to ask more detailed open-ended
questions and follow up questions to elaborate on certain points to capture the user
needs. This method allows the interviewer to take control of the interview and direct it
in a way so they can focus on more important aspects of the proposed system.
This ensures the most important aspects of the system development are identified
within a limited time constraint. One interview tactic interviewers use to determine they
are gathering as much information as possible is to ensure the interviewee feels
comfortable and relaxed in the interview environment. Open ended questions can
be asked resulting in more detailed responses. Follow up questions can be used to
ensure a previous response is fully understood. The interviewer can take notes during
this process to reflect on and analyse the content later (Graham Oakes, G. 2008).
Discovery Prototyping: this involves developing a low fidelity throwaway prototype of

the proposed system. This approach can be useful in situations where the
requirements are vague. This allows the end user to experience and get an impression
of the system and determine whether it meets their needs. Developers will meet with
the user to identify the system requirements and investigate what requirements that
can be established. Once the requirements are identified the prototype is developed
and given to user to assess whether or not it meets the requirements previously
discussed. This gives the user an opportunity to be more involved in the project and
provide feedback on the prototype that will be used to identify the requirements for the
15
final software to be developed. The prototyping approach allows the developer to get a deeper
understanding of the requirements through continuous communication and each prototype iteration or
cycle (Tata McGraw-Hill PublishingInstructional Software Research and Development, 2007).
2.4.1 Chosen Information Gathering Technique

The most efficient way to gather information for this project would be the use of a
Questionnaire, distributed online to the general public. This will reach a large user
group, producing a diverse and varied result set for analysis providing detailed scope
on what users from different age groups and backgrounds are expecting from an
application such as this. It will give valuable insight into the age group, how often the
user utilises online banking, if they would prefer assistance through an advisor on
phone or using a friendly chatbot service, level of computer literacy and their expectations from the
application.
These questionnaires will be completed using Survey Monkey, which provides free
cloud-based expert custom templates for surveys and robust analytical features with
a UI to display data analysis in a number of formats that make it straight forward to read and analyse
the data to identify requirements.
The structured questionnaire will be made up of open and close ended questions as a convenient means
of extracting information from users. There will be clear descriptions throughout the online
questionnaire making it easy to navigate through and use, increasing the likelihood of participation from
users. Clear and concise information will also be given on how the user should respond to each question
accordingly (M. McGuirk , P.M and P. O'Neill, P. 2016).
The questionnaire will be well presented with a consistent colour theme throughout provided by Survey
Monkey. A range of different questioning techniques will be used such as point scales, ranges, open
and closed question. The questionnaire elicit the requirements for the chatbot can be found in the
appendix.
This questionnaire will provide a great understanding on how the public would respond
to using a chatbot to assist them with their banking.
As the purpose of this project is to provide a chatbot to assist with online banking, the question relating
to users who have had past experiences in interacting with a chatbot, see question 5figure 7.1.1 of
appendix A indicates valuable insight into whether or not it was an efficient and enjoyable experience
for the users.
16
It’s clear from analysing the responses of this question that a small number of the respondents have
interacted with a chatbot before as only 11.54 % had previously interacted with a chatbot. Tthat the
majority of users that
have interacted with a chatbot, even if the interaction did not go seamlessly had a very positive opinion
of chatbots in general and actually felt as though they were communicating with
a human as the conversation felt very natural. , Tthis is due to the chatbot having the
ability to understand the natural language used in messages and generate an
appropriate response through various artificial intelligence methods such as natural
language processing. Users found it to be very informative and efficient when replying
to their queries . Although some respondents found that it could take a long time to
reply, or provide very simplistic responses to more complex tasks and become
unresponsive. This reaffirms the need for the chatbot to have a method that will give
the user guidance if it does not understand the context of the conversation or if it
cannot assist the user any further. This may be in the form of a link in the chat
conversation to get assistance from a human or a list of instructions the chatbot can
understand.
The results from the question 4 of appendix A‘Do you feel online banking is secure?’ are displayed in
figure 7.1.2 of appendix outlining there is only a only a 31.028.34% difference in the amount of users
that feel online banking is secure. Out of 264 responses 6558.383% feel online banking is secure
however a lot of users also worry about their personal information being compromised, stolen or
account being hacked and a criminal gaining access to their financial information and committing fraud.
Due to the security concerns most people have in regards to about online banking the chatbot will have
2-Factor Authentication implemented. This feature will reassure users that their account will be
protected, even if a criminal gains access to a user personal information they would need the users
mobile phone in order to successfully access their account as a unique code is sent to the users phone
each time upon requesting to login, the code is validated to allow users access to their account.
2.5 Functional and Non-Functional Requirements

Functional and Non-functional requirements are identified through the analysis of the
data collected from the survey. Functional requirements are the features and
functionality that the system must have or be able to perform whereas non-functional
requirements define the manner or characteristic the system must have such as:
performance, usability, modifiability, maintainability, security, scalability, reliability,
17
availability, configurability and design constraints (Frank Tsui et al,.., 2016)
2.5.1 Functional Requirements

• Allow unregistered users to register on the application and save their details to the database.
• Provide confirmations notifications through emaiSMSl.
• Registered users will be able to login, once login details are submitted to database the user will be
presented with a QR code implemented through Google’s Two-Factor Authentication and a unique
code will be generated and sent to the user’s mobile device.
• The chat bot must allow users to view information about accounts held by them i.e. savings, loans,
current account.
• The Chat bot will allow users to view their transactions through a transactions statement sent to the
users email.
• The Chat bot will integrate with the TrueLayer Starling API which will return data about the users’
bank account.
• The Chat bot will assist users with their queries and carry out appropriate actions such as scheduling
appointments. with finance consultants.
• Users will be able to converse with the Chat bot through voice or text commands and it will
understand what the user is saying through natural language understandinprocessing provided
through the integration of Dialogflow API.
• The chat bot should be able to maintain the conversational state when the context may be unclear
through previous messages and conversations.
• Provide text and audio responses.
2.5.2 Non-Functional Requirements

• The chatbot must be efficient with very little lag in response time for instance no longer than 5
seconds to reply to a user message.
• The chatbot must be reliable with next to no faults or bugs
• The database must be scalable to adopt to a growing number of users
• The chatbot must be secure as sensitive data is being used, Googles 2-Factor Authentication will
be implemented as an extra security feature
• Comply with data protection laws such as the Data Protection Act 1198
• The use of natural language used to interact with the chatbot promotes human computer interaction.
• Provide accurate responsesaccurate responses to input
18
• Appropriate handling of unexpected input & , and correctly inform the user if it cannot provide
assistance
2.6 Software and Hardware Specification

The software and hardware requirements necessary to implement the
chatbot are stated below.
2.6.1 Software Requirements

• Oracle VM VirtualBox
• Vagrant
• Laravel Homestead
• Nginx Server
• MySQL DB
• PHP 7.1
• JavaScript
• Sublime Text 2
• Laravel 5.6
• Dialogflow
• Ngrok
2.6.2 Hardware Requirements

• PC running Linux-Ubuntu to act as server to host chatbot locally
• Processor – intel core i5
• Android Device
2.7 Project Planning

A set of project milestones and deliverables are produced illustrating the project breakdown to create
a management plan to adhere to throughout the project lifecycle, however there maybe slight
changes in each iteration as some requirements may require modification. The steps required in each
iteration are outlined to meet each project deliverable. Produce risk management plan and a project
plan displayed through the use of a Gantt chart showing the projects tasks for each semester.
19
2.7.1 Project Milestones and Deliverables

The project milestones will be used to evaluate the projects progression, setting dates
for tasks to be completed and identifying which objectives have been
satisfied.
Table 1: Project Milestones.
Milestone Milestone Description Completion Date
1 Literature Review 06/11/2017
2 Analysis & Requirements Specification 21/11/2017
3 Complete Interim Report 24/11/2017
4 Design 20/01/2018
5 Iteration 1 26/01/2018
6 Iteration 2 15/02/2018
7 Iteration 3 9/03/2018
8 Iteration 4 –final build 26/03/2018
9 Testing 03/04/2018
10 Evaluation 05/04/2018
Table 1: Project Milestones
The project deliverables are outlined to construe the end result of this project. The deliverables are
shown in table 2 below:
Table 2: Project Deliverables
Deliverable Deliverable Description
1 Interim Report
2 Web based Chatbot
3 Final Report
4 Viva Demonstration
Table 2: Project Deliverables
20
2.7.2 Development Iteration Breakdown

When using the incremental model its vital to establish what functionality is required in each iteration
to lead to the successful development of the project. If any new requirements are discovered during the
development process they will either be added to an existing iteration if small or a new iteration will be
added to the project plan.
Table 3: Project Iterations
Iteration Functionality to be implemented
Iteration 1 • Install and configure VirtualBox VM and Vagrant

• Install Laravel Homestead
• Connect to database
• Add Login functionality using Auth library
• Implement Google Two-Factor Authentication
Iteration 2 • Integrate Starling API

• Have chatbot make calls to Starling API to return customer data when paired
with the appropriate user intent and action
• Integrate Fixer.io API to return conversions when paired with appropriate
user intent and action
• Implement Email transactions
Iteration 3 • Implement the Dialogflow API

• Train Chatbot using intents – expected questions made by users
• Train chatbot to hold a conversation with user
• Create webhook to receive incoming messages from the user and extract
entities
• Implement Twilio API for appointment confirmation
• Implement speech recognition and synthesis
Iteration 4 • Implement Chabot UI

Final Build • Integrate app to google assistant and Google Home
Iteration Functionality to be implemented
Iteration 1 • Install and configure VirtualBox VM and Vagrant
21
• Install Laravel Homestead

• Connect to database
• Add Login functionality using Auth library
• Implement Google Two-Factor Authentication
Iteration 2 • Integrate Starling API

• Have chatbot make calls to Starling API to return customer data when paired
with the appropriate user intent and action
• Integrate Fixer.io API to return conversions when paired with appropriate
user intent and action
• Implement Email transactions
Iteration 3 • Implement the Dialogflow API

• Train Chatbot using intents – expected questions made by users
• Train chatbot to hold a conversation with user
• Create webhook to receive incoming messages from the user and extract
entities
• Implement Twilio API for appointment confirmation
• Implement speech recognition and synthesis
Iteration 4 • Implement Chabot UI

Final Build • Integrate app to google assistant and Google Home
When each iteration is completed and the finished solution developed. The project
will be evaluated to highlight what was learnt and reflect upon the entire project life
cycle. This gives me the opportunity to discuss what was achieved and what steps or
features I would do differently in future projects from what was learnt during this
experience.
2.7.3 Risk Management

Identifying and managing the risks and implications associated with any project is a
vital step in reducing the risks impact on the overall progression of the project.
Identifying the risks implications early on in the project allows preventative actions to
be put in place the minimise the likelihood of the risks occurring. The following actions
22
will be taken to reduce risks in this project; Github to backup code and prevent loss,
use own initiative to keep on top of project plan, regularly make backups of
project report and important documents, if an unplanned day off occurs make up for
work lost.
2.7.4 Project Management

A Gantt chart created in MS Project displaying the project timeline as is shown below in figure 5.4.4.1
Gantt charts are used to identify when the project milestones should be completed so the project is
delivered on time. A breakdown of the tasks, the order in which they will be competed and their time
allocation is shown below.
23
Figure 5: 2.7.4.1 - Gannt chart showing project plan
24
3. Design
3.1 Architectural Design

Figure 63.1 provides a clear and concise abstract architectural design of the proposed chatbot.
25
26
Authenticator
App
Send
SMS
Figure 6:3.1.1 Chatbot Architectural Design.
Users will interact with the chatbot from the web client and Google assistant app for Android
devices. Users will carry out their interaction with the chatbot in the form natural language
voice or text-based phrases. With the Google assistant integration, users will receive rich
responses, such as; images, and cards thus improving ease of use and interaction experience
for users as less typing and effort will be required when interacting with the chatbot mainly
through the use of voice commands. The google assistant integration means the chatbot is
easily be extended by Google Home.
The client side of the application is developed using g HTML5, CSS, JavaScript and Laravel
Blade.
This is a templating engine built into the Laravel framework which follows the MVC
architectural pattern. Laravel Blade compiles and caches view files to improve the
performance of the application. Blade template supports separation of view files to be split up
27
into sections using the ‘@section’ annotation, which enhances the readability of frontend
code, as HTML,PHP and JavaScript can be clearly separated into particular sections of the
view and the ‘@yield' annotation is utilised to render the content for sections of the view.
Blade utilises template inheritance to allow views to derive the base layout of another view to
modify and reuse UI components dynamically without replacing the code in the view it is
inheriting from. Blade also has its own annotations for control structures which may be
required in certain view files, for instance; the ‘@foreach' annotation (Laravel, 2018).
The client side also avails of the chart.js CDN (Content Delivery Service) to display the
user’s spending details in a graphical format so they can easily see where they spend the most
of their money.
The Laravel framework is also used to implement the Web client and Webhook which
receives the JSON data from the NLU via HTTPS POST requests. The Webhook is
implemented as a controller class so the corresponding web route can be is defined as an
endpoint,
where the NLU posts real-time payload to the Webhook. The Webhook receives data from
the NLU when it detects an intent has been triggered by the user from the client-side, Google
Assistant or Home devices.
A webhook receives data posted to it from another service when an update or event occurs
whereas an API follows a RESTful design approach that makes requests to retrieve the data
(elastic.io, 2018).
Dialogflow is utilised as the natural language understanding engine (NLU), which allows the
chatbot to be trained to recognise entities and intents in a user utterance. The Chabot will be
designed using intent mapping on the Dialogflow console to route user utterances to a
collection of phrases.
The Ffigure 73.1.2 represents the flow of data as an intent is invoked by a user: Dialogflow
recognises if an intent was triggered through its training using a collection of typical user
phrases. This data is then posted to the fulfilment webhook. The required entities/parameters
are extracted along with the action and a custom response is generated and returned to the
28
user either visually or spoken from the webhook. The Webhook parses and validates the JSON
from the payload it receives and encodes the JSON response generated to be sent back to the
user for both text and voice interaction.
Dialogflow
NLU Intent Entity Action Fulfilment/ Response

extraction Webhook
Intent detected
Figure 7: 3.1.2 “Intent invoked”.
The Webhook parses and validates the JSON from the payload it receives and
encodes the
JSON response generated to be sent back to the user for both text and voice
interaction.
3.2 Database Design

Laravel also comes with Eloquent ORM, which “provides an ActiveRecord implementation
for working with database objects” (Laravel, 2018). The data is stored locally using a
MySQL Workbench database, it is small-scale to conserve memory. Every table within the
database is represented using a ‘model’ which allows logic and relationships between
database objects to be defined and manipulated.
The database model for the proposed chatbot is outlined in Figure 83.2.1 with defined
multiplicity. The users table contains data about the users of the application and whether or
not they are verified through google 2 Factor Authentication. The appointments table holds
data about the user’s appointments such as ‘topic’. This holds information specified by the
user about what the appointment is for. The ‘booked’ attribute is set to a Boolean value to
determine whether or not the user already has an appointment. The accounts and transactions
29
table holds data regarding the usersuser’s bank accounts, for instance they may have multiple
account types with the one bank such as current and savings, however the users online
banking credentials they possess for each bank are not saved.
Figure 8: 3.2.1 UML Database Model.
3.3 Class Diagram

The class diagram shown in Figure 9 3.3.1 illustrates the classes required to implement the
chatbot. This will act as a guide for implementing the chatbot however other classes may be
identified or modified during development to meet the requirements.
30
31
Figure 9:3.3.1 UML Class Diagram.
3.4 Speech Recognition

Users are able to interact with the chatbot using voice commandsspeak to the chatbot through
the built inbuilt-in microphone on the chatbot device.
The HTML5 speech Recognition API is implemented within the web app for speech
recognition and synthesis.
HTML5
🎤
Speech-To-Text/
Speech Recognition API Text-To-Speech
Figure 3.4.1 “Speech Recognition”
Dialogflow will then take the deciphered speech and convert it into a structured JSON object
it can analyse, this is used as the text input. The JSON response is given an entity matching
score also known as confidence score. This score expresses how well the NLU engine
matched the user input to an intent defined on the console (DialogFlow, 2015). Scores range
between 0-1, 1 being an exact match. Below is a snippet of the JSON object response. User
says: “Who are you?”.
{{ “fulfilment”:
“fulfilment”: {{
“speech”:
“speech”: “I
“I am
am your
your smart
smart banking
banking bot,
bot, use
use me
me to
to get
get your
your account
account balance
balance quickly,
quickly, view
view transactions,
transactions, debits,
debits, account
account
details and receive currency conversions.”,
details and receive currency conversions.”,
“messages”:
“messages”: [[
{{
“type”:
“type”: 0,
0,
“speech”:
“speech”: “I
“I am
am your
your smart
smart banking
banking bot,
bot, use
use me
me to
to get
get your
your account
account balance
balance quickly,
quickly, view
view transactions,
transactions, debits,
debits,
account details and receive currency conversions.”
account details and receive currency conversions.”
}} ]]
},
}, “score”:
“score”: 11 -- Exact
Exact Match
Match
},
},
32
says: “Who are you?”.
3.5 Activity Diagrams

Activity diagrams are drawn to specify the informational flow of the data and outline the
processes that are involved in the chatbot. They represent the users’ behaviour when
interacting with the chatbot and how it responds to these actions through processes. These
will breakdown the distinct functionality of the chatbot.
Figure 10 3.5.1 describes the chatbot logic when a user requests currency conversion. Once the
NLU evaluates the user input and determines a context, the user utterance is passed to the
Fulfilment webhook. The webhook utilises POST endpoints and accepts JSON returned from
the chatbot NLU. The webhook then determines the associated action of an intent the user
supplied utterance is mapped to; the action of an intent was defined in the Dialogflow console
during training. If the action is present it will extract the entities from the given utterance and
call the Fixer.io API and return the appropriate response encoded in JSON format. If the
NLU cannot recognise the intent of the user supplied utterance it will prompt the user for
more information to try and recognise the intent and context to provide fulfilment for the
user.
33
34
Figure 10: 3.5.1 “Currency conversion” UML activity diagram.
35
FFigure 11:3.5.2 “Account balance” UML activity diagram.
36
The diagram shown in Figure 113.5.2 outlines the flow of data when a user requests their
account balance. The NLU determines the context and presence of required entities in user
input that was matched to an intent. The webhook then receives the user input in JSON
format.
Figure 12: “Registration and Login” UML activity diagram.
The webhook is utilised to deliver appropriate custom fulfilment to users by

extracting the action of the intent it receives and checking whether the action is present in the
intent. It then calls the appropriate function based on the current action. If the action is
present the function getBalance() is called. This method calls the TrueLayer API which uses
REST endpoints. The response is then encoded in JSON format and sent to the chatbot NLU.
The Laravel framework comes with an authentication package to handle authentication of the
application. Figure 123.5.3 outlines the process for login and registration. Users must first
register on the application to set up an account. If the user attempts to login before registering
they will not be allowed access the chatbot view. Upon requesting to login, a database check
is performed to determine the validity of the user’s login details. If the details are valid the
user will be redirected to the Two-Factor Authentication view. In this view the user will be
required to scan a QR code displayed on screen to receive a unique code which must be
37
entered to verify their account, if the code is correct the user is redirected to chatbot view.
Figure 12:3.5.3 “Registration and Login” UML activity diagram
3.6 Sequence Diagrams

The sequence diagram shown in Figure 13 3.6.1 highlights the operation of the user requesting
their banking information through the chatbot, outlining the fundamental process involved. The
webhook receives the user input from the Chatbot NLU engine (Dialogflow) in the form of
JSON. The webhook determines the action of the defined intent posted from the NLU by
parsing the JSON it receives. If the action matches a defined banking operation, the webhook
will make the appropriate method method call to the TrueLayerClientStarlingHelper Class,
which calls the TrueLayer Starling API and returns the data to the webhook to generate a for
the response. Once the webhook receives the data returned by the API it encodes the data to
JSON which is as the response sent back to the NLU to be displayed to the user through the
Chatbot view.
38
Figure 3.6.13: “Request banking Information” sequence diagram.
Figure 14:3.6.2 “Appointment & confirmation”
In order to implement a successful chatbot the user group that will be interacting with it is
carefully thought through. The target audience for the chatbot consists of different age groups
with varying technical ability, as mobile banking is relevant across a broad range of age
groups. To appeal to all age groups the chatbot will utilise simplistic and straightforward
responses. Google assistant allows developers to determine a voice for their bot that matches
the personality and utilise rich responses (Actions on Google, 2018). However the majority
39
of users of an application like this already carry out the most part of their online interaction
through some messaging platform, so the use of such an application should come with ease
(Interactions.acm.org, 2017).
3.7 Conversational User Interface Design

Interaction will take place within the chatbot in the form of natural language and this will be
the focus of design in regards to conversational user interfaces (CUI). “Conversation as a
method of interaction is often referred to as the new UI” (Actions on Google, 2018).
(cs.hmc.edu, 2018) define a set of design patterns to adhere to when designing
Conversational User Interfaces.
Table 4: CUI Design Principles.
PRINCIPLES DESCRIPTION
1. Design by personality The bot personality should provide a natural and unique interaction
experience for
users. The personality should be designed to guide the users through the
interaction
which may include being aware of what the user may say or want. The
personality is
the style and manner the bot holds throughout the interaction.
2. Flexibility in response Flexible responses should be provided to user input, the bot should have
various/random responses to the same user input.
3. Text vs Custom It should be clear to the user what input is expected to carry on in the
Buttons converastion. A
fallback response should be used in situations where the input is not

recognised such
as suggesting other operations that are available. Instead of solely relying on

natural
text commands from the user, a good UI should render structured options to
the user
in form of buttons or text.
4. Simplicity in Be concise and clear in response to questions. The conversation flow should
Interaction be
40
straight and not branch out to complex paths.
5. Conversation Flow The conversation flow is important to ensure that the user is not irritated in
situations
where the bot may not be able to provide an appropriate answer immediately.
The
user would then have to undergo a conversation in order to determine the

answer to
their query.
6. Tasks and duty The bot should deliver explicit tasks for a particular domain.
specifications
7. Rigid syntax, then People are prone to spelling mistakes or using colloquial phrases which may
NLP not be
understood by the bot.
8. Empathy and It is important that the bot builds a rapport with the user and convey emotion
emotional state in responses based on certain user input.
9. Keep conversation Users want to interact with the bot to receive answers or reach a goal quickly.
short Avoid long winded conversations as it will make the interaction feel
burdensome. This helps avoid ambiguity.
10. Triggers and actions The bot needs to persuade users and seek ways to encourage users to carry out
specific actions through the bot.
11. Predict and Bots can analyse previous input from users for important parameters, in order
personalise to
anticipate a user’s predisposition. It will then provide an answer that would

be unique
to that particular user.
12. Fully/partially NLP should be used for automated tasks.

automated
13. Providing a way out Allow user to begin the conversation again or exit the conversation.
14. User boredom The conversation should engage the user throughout the interaction.
Table 3: “CUI Design Principles”
A set of design patterns where outlined for domain specific chatbots including those utilised
within a finance industry.
Table 5: CUI Design Patterns.
41
DESIGN DESCRIPTION
PATTERNS
1. Follow-up questions Follow-up questions should be specific in the context of the
conversation. An example of this is when the user wishes to request
a currency conversion. The chatbot would ask; “what is the
amount and currencycurrency you wish to convert to?”
2. Cognitive load Chatbots should present coherent choices to users to reduce the
amount of effort needed for the user to interact with it. This can be
achieved through the use of buttons and images.
Table 4: “CUI Design Patterns”
3.8 Dialog Design

As main focus of interaction is between the user and the chatbot, which is carried out through
natural language, the conversation design is a crucial aspect to consider when designing the
chatbot. To achieve this; appropriate dialog will be designed through the Dialogflow console
using follow-up and fall-back intents. A particular intent will be set, that once invoked by the
user, the chatbot will end the conversation effectively regardless of the conversational context.
The dialog is the collection of words and phrases used to respond to user input. The dialog will
User: “please book me an appointment”
be designed with human like phrases to respond to users with in order to sound more natural.
Another design aspect will include fall-back intents. If a user utterance cannot be matched to
an intent, the chatbot will respond with a message stating it did not match the input to an intent
and prompt the user to repeat the query or provide more details. The dialog is also designed
using follow-up intents, these provide a more natural conversational flow when interacting with
the chatbot. Follow-up intents extract more information from users as and when required, for
instance; a user may say;
User: “Please Book me an appointment”
The chatbot is aware it needs more information from the user such as; date, time and phone
number and will therefore use follow-up intents to extract the required information from the
42
user. Dialogflow recommend designing a linear dialog for interactions where data present in
the conversation is extracted to help the users achieve their goal, this is applied to the
design specification is applied to the chatbot as it is an tasked-based interaction chatbot that
consist of banking domain knowledge. The user utterance can be said in many different ways:
User: “Show me my balance?”
User: “What is my balance?”
User: “Tell me my account balance?”
In this example the chatbot will extract the required entity parameters ‘balance’ and
‘account’ in order to return a custom response to the user (Dialogflow, 2018).
Figure 1453.8.1: “Banking dDialog” Figure 1563.8.2: “Ccurrency

conversion dialog”
The Ffigures 14 5s 3.8.1 and 15 6 3.8.2 illustrate show the design for the Ggoogle assistant
application. It follows a
simplistic layout so it is intuitive and efficient for users, in comparison to other forms of
traditional UI (DZone, 2018). Users can directly ask a question to the chatbot using the input
box or tapping the devices microphone. Once the icon is tapped it will activate the devices
microphone and make the user aware it has started recording and the chatbot response will be
displayed on screen instantly. Google assistant uses specific colours for each message. The
43
use of different colours will be utilised to distinguish between the chatbot response and the
user input.
Figure 167:3.8.3 “Web-based chatbot”
Figure 1673.8.3 outlines shows the design for the web-based chatbot view. Users can see how
much they
spend and what they spend their money on in a graphical format, this will help users identify
over expenditures, for instance; they may realise they spend far too much on coffee each
week. It is made visually clear to the user what the bot’s response is, as each individual
message has an associated title and colour depending on whether or not it was sent by the
user or the bot.
Figure 178 3.8.4 shows the QR code displayed to users for them to scan using the
authenticator app on their device. Users will not be able to access the chatbot view without
verifying their account first as users will be prompted to enter a unique code after each login.
Once the QR code is scanned a unique 6 digit number will be sent to the user’s mobile device
44
for them to enter. Descriptive text will be used in order to guide the user through the process
of downloading the authenticator app.
45
Figure 178: 3.8.4 “Verify Account View”
Figure 189:3.8.4 “Transaction Email”
Users will be able to view their transactions on a regular basis without the need to navigate
through various web-pages. Once the user requests to view their transactions an email will be
sent to their email address that was provided at registration.
Design Conclusion
4. Implementation and Testing

The main functionality of the chatbot is outlined. The chatbot was implemented using the
MVC approach , and adheres to Laravel blade template stated in section 3.1 of thethe design
phasechapter. Chatbot
specific testing is carried out and the results are outlinedanalysed.
4.1 Configuring the development environment:

Laravel homestead was chosenused foras the virtual development environment. This is
essentially a
vagrant wrapper for Laravel. It comes with all the required dependencies to create a Laravel
application, to include; PHP, MySQL, Nginx Server, Node, Git, Ubuntu and ngrok (Laravel,
2018). This saves a lot of time as it avoids the need to manually install any required
packages, in order to set up the ideal development environment for PHP projects. However,
Homestead requires installation prerequisites to include; Vagrant and VirtualBox. Laravel
46
was also added globally to the root directory in order to utilise the framework and create a
project. Figure 19204.1.1 shows the command which was executed used to install the
Laravel\Homestead vagrant box
C:\Users\dana\Documents\bot\bankbot>composerrequire laravel/homestead --dev
in the project directory.

Figure 1920:4.1.1 “”Adding Homestead vagrant box to project”
To configure Homestead, the command presentedshown in figure 2014.1.2 was utilised applied
to initialise the environment inside the project directory.
C:\Users\dana\Documents\bot\bankbot>vendor\\bin\\homestead make
Figure 2014.1.2: “Configure Homestead command”

The homestead command shown depicted in figure 2014.1.2 creates a Homestead.yaml file to
map to the applications folders in the local directory to the appropriate location on the vagrant
box so. Tthe virtual server can then determine the location of the project. Once the
Laravel/Homestead vagrant box is configured, the vagrant box can host the chatbot on an Nginx
server. The project Homestead.yaml file, is shown in figure 212,4.1.3 outlinesing the applications
location on the vagrant box.
Figure 2124.1.3: “Homestead.yaml file”
Note the IP address at the top of the file shown displayed in figure 2124.1.3 is the IP for the
Vagrant box
and the connected database is specified declared under ‘databases:’
To serve the project the “vagrant up” command shown in figure 2234.1.1, this is
executeddeployedrun withinin the
47
project directory through the terminal.
C:\Users\dana\Documents\bot\bankbot>vagrant up
Figure 2234.1.4: “Vagrant up command”
The vagrant box also contains the operating system required for this project which is Ubuntu
Linux. In order to connect to the virtual machine an SSH terminal command is used to
establish a secure connection to the vagrant box.
C:\Users\dana\Documents\bot\bankbot>vagrant ssh
Figure 2344.1.5: “Vagrant ssh command”
4.2 Authentication:
An authentication package is included with the Laravel framework that handles authentication.
The Laravel framework comes with an authentication package to handle authentication. The
command definedoutlined in figure 2454.2.1 shown below generates the foundation for
authentication. This creates the
controllers, views and routes to support authentication.
vagrant@bankbot:~/code$ php artisan make:auth
Figure 2454.2.1: “php artisan auth command”
4.3 Database Development

Once the authentication package was addedinstalled, the database was subsequently migrated
to includeadd the users table
to the database. Laravel comes with built in support for creating migrations to define the
database schema in order to create the required tables. Migrations allow database tables to be
updated or rolled back to a previous version if necessary. The artisan command shown in
figure 256 4.3.1 is used to migrates the database tables defined in newly added migration files.
48
vagrant@bankbot:~/code$ php artisan migrate
Figure 2564.3.1: “php artisan migrate command”
The migration command outlined in figure 267 4.3.2 creates a new migration file to update the
users table to store their Google 2-factor authentication credentials.
vagrant@bankbot:~/code$ php artisan make:migration AddGoogle2faToUsers
Figure: 4.3.267: “php artisan make migration command”
4.4 Google Two-Factor Authentication:

As an extra layer of securityTo further enhance security, Google Two-Factor Authentication
was implemented given the
context of the chatbot. The application uses an authentication package to implement Google
Two-Factor Authentication. The package was installed through composer using the
Following commands:
C:\Users\dana\Documents\bot\bankbot>Composer require pragmarx/google2fa-laravel
C:\Users\dana\Documents\bot\bankbot>Composer require bacon/bacon-qr-code
Figure 27:84.4.1: “G2FA function”
The G2FA() function presented defined in figure 278 4.4.1, displays the code used to
implement 2–Factor
authentication. The authenticated user is obtained and an instance of the Google2FA facade is
created. A unique secret key is added appended to the user’s record and saved to the database.
The
49
getQRCodeGoogleUrl methodfunction of the facçade creates an Image URL for the QR code,
thus
allowing users to scan the QR code using their authenticator app. The users email is sent along
with the secret key, and the application name (BankingBot), to their authenticator appThe users
email along with
the secret key., Users must manually enter the which they are required keyto enter upon
verifying their account. is sent to their
authenticator app along with the name of the application (BankingBot), this is to identify
which key should be used for their account. The QR image URL is then passed to the view to
display the QR code.
Figure 2894.4.2: “verify_g2fa function”
Figure 289 4.4.2 shows depicts the verify_g2fa() controller function which handles verification
of the unique
secret key submitted by the user., This particularthe secret key is obtained from the incoming
request which
holdings the posted data. The window variable specifies the time frame for how longwhich the
key remainsis
valid. A secret key is created every 30 seconds. If the key is valid, a database operation is
performed to set the boolean user attribute g2fa_active to true in the database , save it to the
database and save the verified useradd it to the HTTP
session to determine if the user is verified between session states. and the user is redirected to
the chatbot view.
50
The Laravel blade file, which defines the frontend for the account verification view, is shown
in figure 29304.4.3, the image URLurl for the QR code is used to display the image within a
HTML
image tag. The user is able to submit their secret key via the form input and Laravel comes
with built in support to prevent csrf vulnerabilities. The input is posted to the corresponding
HTTP route outlineddefined in figure 3014.4.4.
Figure 29304.4.3: “Verify Account view”
The web routes used to handle account verification are shown in the below snippet:
Figure 3014.4.4: “Web Routes”
An authorisation link is embedded within the controller class which in turn redirects the user’s
browser to the TrueLayer’s Authorisation Server. A list of real world banks is presented where
the user can select their bank and login with their online banking credentials.
4.5 Open Banking API: :

The TrueLayer API was integrated in order to access the user’s banking data enabling the
chatbot to conform to Open Banking standards. This API allows users to select from a wide
51
range of banks in which they have a pre-existing bank account with. Users can view their
banking information from various accounts from within the chatbot.
After successfully logging into their bank account and permitting the application access to their
sensitive data. TrueLayer API authenticates users through the OAuth2 RFC6749 flow to grant
the application access to the users’ personal banking data on their behalf once redirected back
to the web application. This is achieved by obtaining an access and refresh token through a
TLS encrypted connection, once the user is authorised and redirected back to the chatbot
application. The refresh token is stored in the database, to uniquely identify users upon each
request. The method used to retrieve the access token is shown in figure 3124.5.1
The TrueLayer API was integrated in order to access the user’s banking data enabling the
chatbot to conform to Open Banking standards . This API allows users to select from a wide
range of banks in which they have a pre-existing bank account with. Users can view their
banking information from various accounts from within the web application. TrueLayer API
authenticates users through the OAuth2 flow to grant the application access to the user
personal banking data once redirected back to the web application. This is achieved by
obtaining an access and refresh token once the user is redirected back to the chatbot
application after logging into their bank account using their online banking credentials.
The refresh token is stored in the database, to uniquely identify users upon each request.
The method used to retrieve the access token is shown in figure
52
Figure 3124.5.1: “getAccessToken Method”
The TrueLayer API is A RESTful API which returns JSON that is parsed and stored in the
database. The API is queried using HTTP requests to specified endpoints. Guzzle is
integrated to send HTTP requests to the API endpoints. A HTTPS POST request is sent to
the token endpoint of the Truelayer API containing the code retrieved from incoming request
after the user is redirected back to the web application after successfully logging into their
bank.
Figure 3234.5.2: “getAccounts method”
53
To retrieve the user accounts a userthey may hold with their bank the TrueLayer data APIapi
endpoint is
queried as outlined figure 323., Tthis returns a JSON object containing the account details, for
instance a user may
have a current and
savings account.
The JSON is
then decoded
and the details are
saved to
the database.
FFigure 4.5.334: “getAccountDetails function”

The users account Id isI s appended to the end of the request URL to return specific
information regarding the account specified on the API call. The users IBAN, BIC, Account
Number and Sort code are saved to the database.
54
Figure 3454.5.4: “getBalance function”
An important feature is to retrieve the users balance, to ensure users can instantly view their
account balance information. Figure 345 4.5.4 highlights shows the getBalance() method
function, t used to retrieve the users balance. This function makes a call to the defined endpoint
specified by Truelayer Data API to retrieve a user’s balance information such as available and
current balance. Firstly the result returned form the request are decoded into a PHP array and
then parsed to extract the users balance which is then stored in the database.
Figure 3564.5.5: “getTransactions function”
Figure 3564.5.5 outlines the function which retrieves the user’s transactions from the
55
API for a specific account. The transactions object returned from the API is filtered by the
‘merchant_name’ index as not all transactions returned contained the index ‘merhant_name’,
each transaction is then saved to the database.
4.6 Webhook Development:

The webhook is developed to implement custom fulfilment for the chatbot. Dialogflow posts
data to the chatbot through HTTPS requests. In order to generate a publicly accessible URL,
the tunnelling tool known as
ngrok was used to expose thecreate a tunnel to localhost fulfilment web route endpoint. This is
vagrant@bankbot:~/code$ share bankingbot.test achieved
through the following
command:
The webhook was implemented as a controller class that expose the web route as an
endpoint. The webhook receives the data from Dialogflow NLU and parses the JSON to
extract the entities present the in user utterance to provide a custom response to the user.
Figure 3674.6.1: “Webhook handle function”
The PHP json_decode() function is used to convert the incoming JSON Object from the NLU
into a PHP variable. The code shown in figure 3784.6.2 is a snap shot of the webhook which
returns a custom response when a user requests a currency conversion. Firstly the
appropriate entities are extracted from the request posted to the Webhook then a method call is
invoked to the function that will query the Fixer.io API using the Currency Rates library
installed and the response is returned to Dialogflow in JSON format. The response
is defined for both speech and text in the response array. Each response is sent with the header
Content type: application/json which will display the JSON in plain text to the user through the
chatbot.
56
Figure 3784.6.2 : “Webhook”
In order to implement currency conversion functionality a PHP library called

“CurrencyRates” was integrated into the project using the following command:
vagrant@bankbot:~/code$ composer require ultraleet/currency-rates
Figure 3894.6.3 shows the convert function which implements the library to call the Fixer.io
API.
The required entities are extracted out of the user response and passed as parameters to the
convert function within the webhook and the conversion is returned in a string format.
Figure 4.6.389: “convert function”
57
Figure 3940.6.4: “Webhook balance response”
When the incoming POST request is received by the webhook, the action of the request is
determined in order to return an appropriate response as outlined in figure 3940.6.4. When the
user requests to view their balance, the entity is recognised by the NLU were appropriate
actions were defined for each corresponding entity created within the DialogFlow Developer
console. The if statement within the webhook evaluates whether the requested action of the
user supplied utterance matches the action defined for the balance entity.
Figure 401.6.5 shows the Dialogflow console recognising developer defined entities from
an example user training phrase used to develop the chatbot’s knowledge base.
Figure 401.6.5: "Example training phrase, entity recognition”
Figure 412.6.6: “usersBalance function”
The usersBalance() function illustrated in figure 41function obtains an instance of the

authenticated user, through the Auth
middleware applied to the corresponding route. A collection of the user’s accounts is
retrieved using the ‘accounts’ function defined within the user model. The users available and
current balance are returned to the webhook.
58
The chatbot was trained to recognise custom entities relevant to the banking domain. This
allows the chatbot to recognises the entities in the user utterance to deliver a relevant
response.
Figure 423.6.7: “User training phrases”
The retrieveTransactions() function will retrieve the users transactions from the database
using the transactions() function implemented in the user model. Laravel’s Mail::to function
is utilised to send the transactions email to the user.
Figure 434.6.8: “retriveTransactions() webhook function”
The amount_spent() function is shown in figure 44,54.6.9 which iterates through each
transaction
and adds each transaction item to an array to work out the total. The sum is calculated for
every outbound transaction which is the amount the user spent over the last 7 days.
Figure 445.6.9: “amount_spent webhook function”
An email is sent containing the user transactions for the last seven days using the Mail::to
command as part of the mail facade. An instance of the logged in user is obtained and the
mail is sent using an instance of the Mailable class created RequestTransactions, the
59
transactions array is passed to the mailable class which is used to construct the email.
The following command creates the Mailable class:
.
The mailable class is used to configure and build the email that is sent to the user. The
transactions object is initialised and passed through to the mail view in the build function
shown in figure 456.6.10.
Figure 456.6.10: “Laravel mailable class”
The Blade file used to format the transactions email sent to users is outlined outlined in figure
467.6.11.
It consists of markdown and HTML to display the users transactions in a table format to
replicate the design of a traditional paper based statement. Markdown components are used
60
to render certain UI elements such as buttons and message component which defines the
message to be displayed to users.
Figure 467.6.11: “Markdown Laravel mailable view”
Figure 478:.6.12 “make_appointment function”
The code definedshown in figure 478.6.12 shows the make_appointment() function which
obtains an
instance of the authenticated user in order to make an appointment. The
required parameters used to make the appointment are passed from the webhook using a
61
method call defined in figure 489.6.13. In the make_appointment() function an instance of the
AppointmentReminder class is created in order to send a text notification of their
appointment. The users phone number and appointment details are passed as parameters.
Figure 4.6.1893: “Webhook code snippet”
Figure 49504.6.14: “AppointmentReminder Class”
Figure 49504.6.14 outlinesshows the AppointmentReminder Class which makes use of the
Twilio API to
send SMS messages to users to notify them the confirmation details regarding their
appointment. The Twilio client credentials are set in the ENV file of the project to ensure
they cannot be easily accessed.
62
4.7 Chatbot Web Client

4.7 Chatbot Web Client:
The frontend of the chatbot makes use of the official JavaScript/HTML5 Dialogflow SDK.
The SDK was built upon to add speech synthesis and custom styling using CSS and HTML5.
An Ajax POST request sends the user utterance to the ‘/query’ endpoint of the Dialogflow
API as shown in figure 501 4.7.15 to query the user utterance, where it is matched to an intent
and
sent to the webhook to supply the user with a response. The Webhook endpoint is defined
within the Dialogflow developer console which uses entity recognition to extract meaning
from text among other AI and ML methods.
Figure 4.7.1501: “AJAX post request to Dialogflow API”
Figure 5124.7.1: “Speech Synthesis function”
Figure 5124.7.1 shows the function that will produce a speech synthesis of the bots response
to
the user in order to provide audio responses reducing the amount of time required to interact
63
with the chatbot. The HTML5 Speech API is implemented for speech recognition and
synthesis. Dialogflow comes with support for integration of the Google Assistant, supported
across an array of devices including google Home and Android devices. Dialogflow allows
developers to extend the Actionss on Google (AoG) console to test the chatbot across multiple
devices. The Google assistant integration is shown below.
4.8 Git Version Control
Figure 523: “GitHub master branch”
Throughout the development lifecycle GitHub was integrated as the chosen method of version
control for the project. As the application was updated during each iteration, and these changes
were committed to the GitHub repository to ensure the codebase for the application was
managed effectively. Figure 523 outlines the use of the Git Bash terminal that was used to push
changes to the master branch of the project.
5. Testing:
A crucial part of any software development lifecycle is testing. This involves carrying out
certain procedures and operations to understand the limitations of the software. It is evident
that with testing the constraints of the application that particular bugs and errors are picked
up and documented through test cases. This will improve the overall standard and quality of
the chatbot and enhance the user experience.
Various testing methods were carried out in order to measure the overall effectiveness of the
chatbot. The dialog was tested to measure the efficiency of the chatbot which includes
measuring how well the chatbot can understand a user supplied utterance, even if miss spelt.
Identifying if the intent was recognised, with average response times between text and voice
64
interactions are also included in said response. The chatbot was tested in a specific manner to
record the performance metrics.
5.1 Simulated Conversational Testing

The dialog was tested to measure the efficiency of the chatbot which includes measuring how
well the chatbot understood can understand thea users supplied utterance for both voice and
text interactions, even if miss spelt. Identifying if the intent was recognised, with average
response times between text and voice interactions are also included in said response. The
chatbot was tested in a specific manner to record the performance metrics.
A Dialogflow command line tool called dialogflow-cli , written in JavaScript, was installed
into the project directory for the purpose of undertaking these tests. This was employed to
replicate how users would interact with the chatbot and metrics were recorded for each
simulated interaction.
To initialise the mock interaction for testing the tool fires a welcome event to the Dialogflow
API once instantiated in the command line, it then and displays the response returned from the
API as the result returned through the
command prompt.
Each type of use case was tested through the command line and the response was returned in
JSON format.
A total of 20 simulated utterances were tested and the results are shown in table 5. Each
utterance tested was uniquely phrased including variations of the same utterance to
accumulateaccumulate a range of results so and thoroughly classify the chatbots understanding
can be thoroughly classified.
From the results shown in the table 65 it can be determined how well the chatbot understood
the
sentence and typical response times are displayed to measure the overall performance of the
chatbot. This also outlines how it handles unexpected input, such as misspelt words. The
cconfidence score is returned along with the intent the test user phrase was matched to.
65
Table 6: “Simulated User Phrases: Test Results”.
Simulated User Utterance Confidence Matched Unrecognised Response Time
(including misspelt phrases) Score Intent(s) Intent(s) (ms)
Hello bankingbot, what is my 1 ✓  3480
balance?
Hello bankingbot, please show me my 0.9999 ✓  3414

balance?
Display my blance 1  ✓ N/A
Show me my account details 1 ✓  3446
please?
“accont dctails” (account 1 ✓  2181
details)
Can I see my account details 0.8700 ✓  3480
How much money have I 0.8299 ✓  3649
spent this week?
What hove I spunt 0.3000 ✓  3308
Wat have I spnt this weke 0.3300 ✓  2504
Can you send me my 1 ✓  7339
transactions for the last seven
days?
tamsucyions 1  ✓ N/A
how much is 5000 Canadian 0.8900 ✓  3228
dollars in Japanese yen
Convert 600 euro to Japanese 0.9499 ✓  3182
yen
and 700 euro 1 ✓  1613
and 800 pounds 1 ✓  3381
what about 500 US dollar 1 ✓  3128
Can ypu please book me an 0.6200 ✓  3380
appointment please?
buuk me apptmnt 1 ✓  3345
66
Book me an appqentmttt for 1 ✓  3322
sqivings on the 23/04/2018
Ok, my phone number is 1 ✓  4212
089456787642
Table 5: “Simulated User Phrases: Test Results”
Table 65 displays the results from simulated "typical" interactions with the chatbot through the
command line tool. The simulated phrases were derived from the user questionnaires and
observing the users interacting with the chatbot, the dialog for the chatbot was devised during
the design stage. Every utterance is given an "Confidence Score", rated on a scale of 0.0 - 1,
with 1 being a complete match and 0 meaning the utterance was not successfully matched to
any intent. However, the NLU has also given words or utterances, where no intent was
recognised, a complete confidence score of 1 as it was matched to the default fall-back intent
to handle unexpected input from the user with ease. It is clear that the response times do vary
depending on the underlying functionality required for the user to achieve their goal, for
instance: ; requesting their transactions via email or the chatbot sending text confirmation of
their appointment.
Overall, the chatbot is able to match intents with the majority of the utterances with only 10%
of the utterances failing to match. This is taking spelling mistakes into account also. The total
number of successful matched intents out of 20 user phrases amounted to 80%. These results
indicate that the chatbot can understand most phrases including spelling mistakes reflecting the
quality of the dialog. The phrase "wat have I spnt this weke" (meaning "what have I spent this
week") obtained a low confidence scored of 0.33 a perfect 1 , however the chatbot was able to
understand the intent of the utterance and identify the action on the understanding score as
shown in figure 534re… even though
misspelt.
67
Figure 54 :534: “Matched Intent to misspelt phrase”
Table 7: “Average Test Results”.
Average Understanding Score Average Response Time Dialog Duration
17.7897/20= 0.8894 60594 (ms) 60.594(sec) / 20 = 3.02 sec 3 minutes and 26 seconds
Table 6: “Average Test Results”
The average response time has been rounded to 3 seconds, which is quite a rapid response
time. This suggests that users will be able to perform their necessary tasks through interacting
with the chatbot relatively quickly which is also reflective onin the overall dialog duration as
shown in table 76, this represents the overall interaction time. The results will be further
elaborated upon in the Evaluation, with user testing results in contrast to the simulated
resultsphrases.
Both the Google Assistant and Home agent were tested through the AoG Simulator on
the Actions on Google console (AoG). The Google Assistant integration was tested through
the simulator using text input while the Google Home was tested through voice interaction on
the AoG simulator. The simulator returns a JSON object holding information regarding the
chatbots understanding level identifying whether the utterance was matched to an intenta
phrase. Similar phrases were used across devices to outline any variations in the NLU
understanding.
68
Table 8: “Simulated Device testing”.
Intent Google Assistant Matched Google Home Matched Web APP Matched
(Text) Intent(s) (voice) Intent(s) (Voice) Intent(s)
1. Correctly spelt ✓ Correctly Correctly

phrase: pronounced phrase: ✓ pronounced phrase:
1.What is my balance 1.What is my balance 1.What is my balance ✓
Incorrectly spelt Incorrectly Incorrectly

phrases: pronounced phrases: pronounced phrases:
1.What is my balan  1.What is my balan 1.What is my balan
2. Wat is my bolonce 2.Wat is my bolonce 2.Wat is my bolonce

✓
Detected as: Detected as:
✓ ✓
1.What is my balance 1.What is my balance
✓ 
2.What is my balance 2.What is my bowl
2. Correctly spelt Correctly Correctly

phrase: ✓ pronounced phrase: ✓ pronounced phrase: ✓
1.Show me my 1.Show me my 1.Show me my
transactions transactions transactions

1.Shw me mi  1.Show me my 1.Show me my

tronsoctions Tronsoctions tronsoctions
1.Show me my ✓ 1.Show me my ✓
Transactions Transactions

phrase: pronounced phrase: pronounced phrase:
1.Convert 200 1.Convert 200 1.Convert 200 ✓

✓ ✓
Autralian dollars to Autralian dollars to Australian dollars to
pounds pounds pounds
69

1. Convert 200 pound to  Convert 200 pound to Convert 200 pound to

yoyo yoyo yoyo
Convert 200 pond to Convert 200 pound to

✓ ✓
euro euro
Intent Google Assistant Matched Google Home Matched Web APP Matched
(Text) Intent(s) (voice) Intent(s) (Voice) Intent(s)
1. Correctly spelt ✓ Correctly Correctly

phrase: pronouncedspelt ✓ pronounced spelt
1.What is my balance phrase: phrase:

✓

phrases: pronounced spelt pronounced spelt
1.What is my balan phrases: phrases:


2. Wat is my bolonce 1.What is my balan 1.What is my balan
✓ ✓
2.Wat is my bolonce 2.Wat is my bolonce
✓ 
✓✓
✓ ✓
2.What is my balance 2.What is my bowl


phrase: ✓ pronounced spelt ✓ pronounced spelt ✓
1.Show me my phrase: phrase:
transactions 1.Show me my 1.Show me my

transactions transactions
70

phrases: pronounced spelt pronounced spelt
1.Shw me mi phrases: phrases:


tronsoctions 1.Show me my 1.Show me my
Tronsoctions tronsoctions

✓ ✓
1.Show me my 1.Show me my
Transactions Transactions

phrase: pronounced spelt pronounced spelt
1.Convert 200 phrase: phrase:

✓ ✓ ✓
Autralian dollars to 1.Convert 200 1.Convert 200
pounds Autralian dollars to Australian dollars to
pounds pounds

phrases: pronouncedspelt pronounced spelt
1. Convert 200 pound to phrases: phrases:


yoyo Convert 200 pound to Convert 200 pound to
yoyo yoyo

✓ ✓
Convert 200 pond to Convert 200 pound to
euro euro
Table 9: “Overall accuracy comparison of Devices”.Table 7: “Simulated Device testing”
Google Assistant Google Home Web App
(HTML5 Speech API)
% Response Accuracy
% Matched Intent 57.14% 100% 85.71%
Thorough testing was performed across multiple simulated devices.

During the testing across multiple simulated devices Iit was determined that the Google
Home device could recognise the intent of an utterance even when colloquial phrases were
produced. It was also identified that the Google Home device could recognise utterances where
71
particular words were incomplete or missmispronouncedpelt as shown in table 87. It was found
that input text provided to the Google Assistant on android phones, follows a more rigid syntax
as it expects the intent present in the given input text to match the defined intent used within
the training phrases, however this only applies when the use of input text is provided by the
user, as the same speech recognition and machine learning algorithms are implemented with
the Google Assistant regardless of the device or surface its embedded in.
The Google
Assistant follows a more rigid NLP syntax, however this only applies when the use of input
text is provided by the user, as the same speech recognition and machine learning algorithms
are implemented with the google assistant regardless of the device its embedded in. The
speech recognition software in the Google aAssistant transforms the users speech, even if
incorrect, to the appropriate input text before itsit’s sent to the NLU. Consequently the Google
Home device This is why
the Google Home device was able to recognise and detect incomplete or missmispronounced
pelt words, as interaction only occurred through voice commands. Based on these results it is
clear that voice interaction is the optimal method of engaging with the chatbot due to it being
able to comprehend pronunciation errors and interpret them accordingly. It was also found
that the Google Assistant could only match an intent of a to a misspelt utterance when the
trained intent or recognised entity was typedspelt correctly during text interaction.
It is clear from the results shown in table 9 that the Google Home device was the prime mode
of interaction with the chatbot, as it was able to match the intent of each supplied utterance,
achieving a perfect score of 100%. Establishing that spoken natural language conversational
interfaces perform best, compared to that of text-based interactions as the Google Assistant
achieved a result of 57.14%. There is a stark contrast between the two conversational interfaces,
implying spoken dialog is a much better method of interacting with natural language
conversational interfaces. The speech recognition on the Google Assistant has a higher
recognition rate in comparison to the HTML5 Speech API as it had an average speech
recognition of 85.71%, suggesting the Google platforms will perform better in real interactions
with users.
Table 10: “Conversation exit rate”.
Intent (n)Turns before exit
72
1. 3
2. 4
3. 4
Table 8: “Conversation exit rate”
Table 10 8 outlines the number of turns the chatbot took to try and understand the intent of an
utterance corresponding to the misspelt phrase defined in table 87 7 before exiting the
conversation. On each turn the chatbot responded with fallback intents such as: "Sorry, I did'nt
quite get that, could you repeat that please?" or "Sorry, I’m having difficulty understanding
what you said, could you say that again?”. This feature handles unexpected input or when the
chatbot cannot understand a given utterance. This also outlines the use of different fallback
intents used even if the user utterance was the same on each turn. The chatbot exited the
conversation due to repeatedly prompting the user for an appropriate utterance and lack thereof,
its inability to understand the intent of the given utterance to successfully match it to a trained
intent to carry out the appropriate action and fulfil the user’s goal.
5. 2 Webhook Testing
To test the custom fulfilment provided through the Webhook, a tool called Postman was
utilised to send HTTPS POST requests to the Webhook. This was used to test the response
sent back from the Webhook to a given user utterance. The body of the request is obtained
from the Dialogflow developer console which is sent with the POST request to the Webhook.
The body holds the JSON object which represents the users phrase along with other
parameters returned by the NLU as shown in figure 545below.
73
Figure 545: “Webhook POST requesttest”.
Postman was used to test the Webhooks functionality in a separate environment, in which it
was developed, meaning errors could be rectified before integrating the agentit with other it
the various various
devices such as: ; Google Assistant and, Google Home or the web based chatbot. This identified
network and access errors as status codes are returned with each response. An example of the
JSON response object returned by the Webhook is shown belowin figure 556XXXXXXXXX:
Figure 556: “Webhook response”.
The integration of other API’s implemented within the project were also tested, to determine
whether it could successfully call the TrueLayer API and identify any required parameters or
authentication headers that were required to be sent with each subsequent request.
5.31 User Testing:

It was decided to carry out user testing across all three communication channels of the
74
chatbot; Ggoogle Aassistant, Hhome and the web-based chatbot. This will give an insight into
the true
quality and overall usefulness of the chatbot which is measured by the user’s interaction
experience. In order to conduct user testing, a set of questions were constructed to evaluate
the chatbot. These questions were distributed as an online questionnaire across a group of
users which can be found in the appendix section of the report.
65. Evaluation
The purpose of evaluating the software is to identify the quality of the chatbot by outlining
the performance attributes and analysing the results, future work is purposed along with a
reflection of the current work completed.
The results from the user questionnaires gave great insight into the overall response
and precision rate. The questionnaires were distributed across a user group consisting of 15
individuals with varying technical knowledge. These were completed by parents and
prospective students at the Ulster University Open Day as part of the school of Engineering
and Computing, this user group is truly reflective of the target audience that would benefit
from an application like this. The results from the user questionnaires will be compared to the
simulated user interactions identifying subjective and objective metrics.
During the questionnaire the users were observed to capture free standing information
regarding the interaction experience which also qualify as subjective measurements. This
allowed the identification of other quality metrics which were not initially considered and
lead to a deeper understanding of the chatbots performance. (How often the chatbot repeated
itself, ).
The metrics shown in table 9 were They completed a user questionnaire targeted to capture the
quality of the chatbot.
The Subjective measurement (n)Users Objective metrics (n)Users subjective
and Naturalness 44% Speech Recognition 80% objective
metrics Likeability 93.33% Response accuracy 86.67% outlined in
table 11 Ease of use 86.66%
were
obtained using a
75
range of question types including Likert scale and numerical scales to accurately assess the
chatbots performance.
Each metric outlined in table 11 was measured in relation to a specific question to gather
qualitative results. These questionnaires can be found in appendices.
Table 11: “Subjective & Objective Measurements obtained from User questionnaire”.gathered from
Subjective metrics (n)Users Objective metrics (n)Users
Naturalness 44.00% Speech 86.67%80.00%

RecognitionSpeech
Recognition
Likeability 93.33% Response accuracy 86.67%
Ease of use 86.66% Speed of 73.33%

InteractionResponse
Table 9:
rate
Speech Recognition 80.00%
“Subjective & Objective Measurements obtained from User questionnaire”
As outlined in section 1.2 the majority of banks struggle to get their customers to use the
technology
in place due to low user satisfaction when interacting with their services. From the results
shown in table 119 it can be established that integrating a chatbot into their banking services
would greatly increase the rate of user satisfaction. This is evident as 73.33% of all users
stated the chatbot was “extremely ” quick comparedfound the chatbot to be a lot quicker tohan
the current technology offered by their
bank, for instance, online banking applications. From completing the user testing phase, it was
found that 86.66% of all respondents foundstated found the chatbot was
either “extremelyvery easy” or “moderatelyeasy” easy to talk to and it achieved an overall
likability of 93.33of 93.33% as outlined in table 119,. Rreaffirming
that chatbots can successfully engage users and increase satisfaction. The objective metrics
obtained from the questionnaire results identify the chatbots performance and quality in a real
interaction. Overall the chatbot received a very high speech recognition rate as 80.00% of users
stated they were “very strongly” well understood by the chatbot. This measurement reflects the
accuracy
76
of the emulated actions during testing as shown in table 7 of testingThe chatbot obtained an
overall understanding score of 0.8894 (89.00%) in the range between 0.0-1 during the
simulated tests as depicted in table 6 of section 5. This result is on the higher tier of the
understanding scale.
Evidence from the results gathered during user testing, through measuring the response
accuracy to real user interaction, that the chatbot has a high competency level as both the results
from the simulated and user tests have a very low variance of 2.33% in regards to the
understanding score. These measurements reflect the accuracy and thoroughness of the
emulated actions performed during testing as illustrated in section 5.
Figure 567: “Users preference”.
It is clear from the questionnaire results that the majority of users preferred to interact with
the chatbot through voice commands, with 66.67% of users preference being voice
77
interaction, as shown in figure 567 . This supports that more users are wishing to be able to
freely
interact with applications and not be limited to traditional modes of interaction. This feature
also offers users the ability to interact with chatbot without it requiring much attention on
focusing on the task at hand, as users ask the chatbot a question and it will recognise the users
voice negating the need for the user to type.
Figure 578: “Conversation flow”.
The focus of any chatbot is to allow users to complete tasks with ease using natural language.
It i’s important that the chatbot clearly understood the user and the user felt confident in
interacting with the chatbot. Its outlined in figure 578 that this was achieved as only 13.33%
of
users stated they were “often” left in a state where they were unsure how to carry on the
conversation. These results reflect low frustration rates among users as 60% of user never
experienced this and 26.67% stating it did not occurred “Not often.” often. This clearly suggests
the
chatbot is able to sufficiently guide users through the conversation with ease providing
appropriate responses to the majority of user phrases.
78
Figure 589: “Chatbot Conversational context”.
Enabling the chatbot to be context aware provides a great advantage for users as they can
simply refer back to a previous conversations and still be understood by the chatbot. This is
particularly useful in long conversational flows, reducing the amount in which the user has to
repeat themselves thus shortening the time taken to achieve a task.
65. 1 Reflection
The results gathered throughout the testing phase support the research conducted within section
1.3, which outlines the prevalence of chatbots within industries, particularly within the banking
and financial sector. The findings prove that the chatbot is a very suitable method for adopting
technology as a means to distribute a service. The developed chatbot allows users to interact
with their bank through natural language interaction, granting users more convenient and
efficient access to their banking information. The project gradually evolved and progressed
throughout its entirety and the main objectives and requirements have been met as outlined in
section 2.5.1 and 2.5.2. This chatbot allows users to connect to various banks they may have
an existing account with all from within the one communication channel through natural
language.
65.1.2 . Future Advancements

One of the integration s streams that were initially thought of during development was to
integrate the chatbot with the Facebook Messenger service, however during development it
79
was brought to light through of recent media events such as the Cambridge Aanalytica
discreditscandal, that which resulted in Facebook haltedpausing their process of allowing other
apps to integrate with its messenger service app review, so this integration was not developed
as Facebook hads stopped allowing developers to create new apps or chatbots through the
service. Although Facebook have now just recently reopened their app review processs, now
allowing developersallowing developers to integrate with the Messenger service again,
enabling this integration feature to be developed for future protypes (Facebook, 2018).
Integrating Google authentication would have been an advantageous feature to implement,
allowing users to link their Google account across the multiple devices the agent is
distributed on, for instance, Google Assistant enhancing the cross-platform experience for
users.
User can find out financial information about their account, such as how much they spent
within a week. From this feature it was thought that during the later stages
of development users should be able to directly query the chatbot to determine where and
what they spend their money on. Although users can view this information in a graphical
format onin the web app it would also be useful if they could find this information out through
interaction with the chatbot.
The chatbot could also be developed to be multilingual. Supporting multiple languages would
help aid banking exclusion in developing countries and enhance the overall accessibility of
technology within the banking industry whilst targeting a larger user group.
From observing the users interact with the chatbot during user testing, there were numerous
suggestions to integrate the chatbot across other popular platforms such as Amazon Echo or
Dot, this would increase the availability of the chatbot and the integration of the DialogFlow
API allow the chatbot to be easily exported Amazon Alexa.
65.2 . Conclusion
The rise and popularity of chatbots is clearly outlined in figure 1 of section 1.3. With this in
mind the data gathered from testing the chatbot justifies the recent growth and demand for
companies wanting to integrate a chatbot. It was determined that chatbots perform at a very
80
high standard and provide reliable and rapid responses to users compared to that of traditional
methods. The average time spent interacting with the chatbot is very low as it provides an
efficient way for users to manage their banking. The low interaction time reflects the high
understanding and speech recognition rates, offered through the adoption of conversational
user interfaces thus allowing users to freely interact with the chatbot to meet the demands of
modern life. The chatbot has proven to fulfil the demand of users wanting instant access and
availability information and services.
Integrating sentiment analysis to capture the user’s emotions. The user’s transactions could
be analysed using NLP techniques to identify spending habits.
Due to the increased development of the IoT, it would be good to branch out into this area as
chatbots produce a
mass amount of data which is needed for concepts such as IoT and Big Data to be successful
which also go in hand with further advances in AI and Machine learning. The
IoT could be implemented through the Google Home device.
81
1
6. References
(United kingdom online banking (2002). MarketLine, a Progressive Digital Media business, 2017)
(Accessed 10/10/2017)MarketLine, a Progressive Digital Media Business. (2002). ‘United Kingdom
online banking’. Available at: http://www.marketlineinfo.com (Accessed 10/10/2017).
Ling, G., Fern, Y., Boon, L. and Huat, T. (2016). Understanding Customer Satisfaction of Internet
Banking: A Case Study In Malacca. ‘Procedia Economics and Finance’, 37, pp.80-85.n (Accessed
11/10/2017)
Barty, J. and Recketts, T. (2014). 'Promoting Competetion in the UK banking Industry', Promoting
Competetion in the UK banking Industry. BBA Competition Report., , pp. 4. Available at:
https://www.bba.org.uk/wpcontent/uploads/2014/06/BBA_Competition_Report_23.06_WEB_2.0.pdf
(Accessed 10/10/2017).
Aburub, F., Odeh, M., & Beeson, I. (2007). ‘Modelling non-functional requirements of business
Processes’. Information and Software Technology, 49(11–12), 1162-1171. Available at:
http://dx.doi.org/10.1016/j.infsof.2006.12.002 (Accessed 10/10/2017).
Cardline. (2010). ‘Half of UK internet users now bank online’, 10(3), pp. 32-32. Availablle at:
http://web.b.ebscohost.com/bsi/detail/detail?vid=1&sid=23ade416-2f14-4608-a8a1-
3de95541175b%40sessionmgr102&bdata=JnNpdGU9YnNpLWxpdmU%3d#AN=47659678&db
=bth (Accessed 10/10/2017).
Half of UK internet users now bank online.(2010). Cardline, 10(3), 32-32, .(Accessed 10/10/2017)
HDFC bank, niki.ai tie up for chatbot banking.(2017). FRPT- Finance Snapshot, , 24-25.( Accessed
11/10/2017)
FRPT Research- Finance Snapshot. (2017). ‘HDFC Bank, Niki.ai tie up chatbot for banking’, pp
24-25. Available at: http://web.b.ebscohost.com/bsi/pdfviewer/pdfviewer?vid=8&sid=c34ab15e-

2256-41a5-8757-f8ef87bd08e8%40sessionmgr104. ( Accessed 11/10/2017)
Ling, G. M., Fern, Y. S., Boon, L. K., & Huat, T. S. (2016). Understanding customer satisfaction of
internet banking: A case study in malacca. Available at doi:https://doi.org/10.1016/S2212-
5671(16)30096-X
2
Hugh J. Watson, H. J. (2017). 'Preparing for the Cognitive Generation of Decision Support', MIS
Quarterly
Executive, 16(3), pp. [Online]. Available

at: http://www.misqe.org/ojs2/index.php/misqe/article/viewFile/705/465 (Accessed: 10/10/2017).
Daniëlle Duijst, D. (2017). 'Can we Improve the User Experience of Chatbots with
Personalisation?', University Of Amsterdam, (), pp. 3 [Online]. Available
at: 10.13140/RG.2.2.36112.92165 (Accessed: 11/10/2017).
United kingdom online banking(2002). MarketLine, a Progressive Digital Media business.
Watson, A. (2010). How to succeed with NLP: Go from good to great at work using the power of
neuro-linguistic programming. Oxford: Capstone .(Accessed 11/10/2017)
KingING, W. B. (2017). Year of the chatbot: Credit unions gearing up for artificial
intelligence. Credit Union Journal, 21(4), 18-18 . (Acccessed 11/10/2017).
Dole, A., Sansare, H., Harekar, R. and Athalye, S. (2015). [online] www.ijettcs.org. Available at:
http://www.ijettca.org/Volume4Issue5(2)/IJETTCS-2015-10-09-16.pdf [Accessed 11 Oct. 2017].
Intelligent chatbot for banking system. (2017). [ebook] Available

:http://www.ijettcs.org/Volume4Issue5(2)/IJETTCS-2015-10-09-16.pdf [Accessed 11 Oct.
2017].
D. Latimore, D I. . Watson, I. & C. Maver. C. (2000). The customer speaks: ‘3300 Internet users tell
us what
they want from retail financial services’.
Yusuf Dauda, S.Y., & Lee, J (2015). ‘ A conjoint analysis of consumers‫ ׳‬preference on future online
banking services, Nigeria: Elsevier Ltd.
Goncalo Bapista, G. and Tiago Oliveira, T. (2015). ‘Understanding mobile banking: The unified
theory of
acceptance and use of technology combined with cultural moderators’, : Elsevier Ltd.
3
Ajimon George. A and G.S. Gireesh Kumar, G. (2013) 'Antecedents of Customer Satisfaction In
Internet Banking: Technology Acceptance Model (TAM) Redefined', Global Business Review, 14(4),
pp. [Online]. Available at: http://journals.sagepub.com/doi/10.1177/0972150913501602
Available
at: https://smartech.gatech.edu/bitstream/handle/1853/58516/evaluation_of_modern_tools_for_an_o
mscs_advisor_chatbot%281%29.pdf?sequence=1&isAllowed (Accessed: 21/10/2017).
Eric Leenders, E. (2018).’ () Banking’s digital revolution continues’, Available

at: https://www.ukfinance.org.uk/bankings-digital-revolution-continues/ (Accessed: 12/10/2017).
Susan Etlinger, S. (2017). ‘The Conversational BusinessTHE CONVERSATIONAL BUSINESS: How

Chatbots will reshape Digital Experiences’. [Online]. Available
at: https://andyblackassociates.co.uk/wp-content/uploads/2015/06/ConversationalBusiness_FINAL-
1.pdf (Accessed: 16/10/2017).
YY Onufreiv, YY. (2017). ‘The Rise In Chatbots Trends’THE RISE IN CHATBOTS TRENDS. THE
RISE IN CHATBOTS TRENDS. - (-), 82-83. (Accessed 16/10/2017)
Eric Gregori, E. (2017). 'Evaluation of Modern Tools for an OMSCS Advisor Chatbot'. [Online].
Available
at: https://smartech.gatech.edu/bitstream/handle/1853/58516/evaluation_of_modern_tools_for_an_o
mscs_advisor_chatbot%281%29.pdf?sequence=1&isAllowed (Accessed: 21/10/2017).
Erik Cambria, E., Bebo White, B. (2014) 'Jumping NLP Curves: A Review of Natural Language
Processing Research', IEEE Computational Intelligence MAgazineCOMPUTATIONAL
INTELLIGENCE MAGAZINE. [Online]. Available at: http://sentic.net/jumping-nlp-
curves.pdf (Accessed: 21/10/2017).
Daniel Jurafsky, D. and James H. Martin, J. H. (2017). An Introduction to Natural Language

Processing,
Computational Linguistics, and Speech Recognition. Third Edition [Online].
Available at: https://web.stanford.edu/~jurafsky/slp3/ed3book.pdf (Accessed: 21/10/2017).
https://www.bba.org.uk/wp-
content/uploads/2014/06/BBA_Competition_Report_23.06_WEB_2.0.pdf
4
Rohan Kar, R. and Rishin Haldar, R. (2016). 'Applying Chatbots to the Internet of Things:
Opportunities and Architectural Elements',, , (), pp. [Online]. Available
at: https://arxiv.org/abs/1611.03799 (Accessed: 06/11/2017).
Nat Landry, N. (2011). ‘ Iterative and Agile Implementation Methodologies in Business Intelligence
and
Software Development’, : Lulu Enterprises.
A B M Moniruzzaman, A. B. M. and Dr Syed Akhter Hossain, S.A. (2013). ' Comparative Study on
Agile software development methodologies', Global Journal of Computer Science and
Technology, 13(7), pp. [Online]. Available
at: https://arxiv.org/ftp/arxiv/papers/1307/1307.3356.pdf (Accessed: 08/11/2017)
Asif Irshad Khan, A.I., Rizwan Jameel Qurashi, R. J and Usman Ali Khan, U. A. (2011). 'A
Comprehensive Study of Commonly Practiced Heavy and Light Weight Software Methodologies
', IJCSI International Journal of Computer Science Issues, 8(4), pp. [Online]. Available
at: https://arxiv.org/ftp/arxiv/papers/1111/1111.3001.pdf (Accessed: 08/11/2017)
Jonathan Lazar, J. (2001) User-centered Web Development, United States of America: Jones and
Bartlett Publishers .
Graham Oakes, G. (2008). ‘Project Reviews, Assurance and Governance’, United Kingdom: Gower
Publishing Limited.
Instructional Software Research and Development (2007) Structured System Anal And Design Isrd,
New Delhi: Tata McGraw-Hill Publishing Company Limited.
Pauline M. McGuirk, P. M. Phillip and O'Neill, P. (2016). 'Using questionnaires in qualitative human
geography', University of Wollongong, Faculty of Social Sciences, (), pp. 11 [Online]. Available
at: http://ro.uow.edu.au/cgi/viewcontent.cgi?article=3519&context=sspapers (Accessed:
12/11/2017).
Melanie Exner-Stöhr , M., Alexander Kopp , A., Leonhard Kühne-Hellmessen , L., Lukas Oldach ,
L., Daniela Roth, D. Aand Alfred Zimmermann, A. (2017). 'The potential of Artificial Intelligence in
academic research at a Digital University', Digital Enterprise Computing, (), pp. 3 [Online]. Available
at: https://dl.gi.de/bitstream/handle/20.500.12116/120/paper05.pdf?sequence=1&isAllowed=y (Acce
ssed: 16/11/2017).
5
Kazuo Yano, K. (2017). ‘How artificial intelligence will change HR’.40(3), 43-44.(Accessed
16/11/2017)
Stuart J. Russell, S. J and Peter Norvig, P. (1995). Artificial intelligence A modern approach.
Noor Atikah, N., Amira Fauzi, A., Rohayanti Hassan, R., Zuraini Ali Shah, Z. A. and Zalmiyah
Zakaria, Z. (2016) Extraction of Non-Functional Requirements in Reviewing Requirements
Specification Document, 2, pp. 1 (Accessed 19/11/2017).
Frank Tsui, F., Orlando Karam, O. and Barbara Berna, B. (2016). ‘Essentials of Software
Engineering’. 4th Edition [Online]. Available
at: https://books.google.co.uk/books?hl=en&lr=&id=fV-
HDQAAQBAJ&oi=fnd&pg=PP1&dq=functional+and+nonfunctional+requirements+definition+soft
ware+development&ots=P3-
Cil3TVN&sig=kWwrD0XWLzWyxfi49W4XVfWxXWg#v=onepage&q&f=false (Accessed:
19/11/2017).
Milan van Eeuwen, V. M. (2017). 'Mobile conversational commerce: messenger chatbots as the next
interface between businesses and consumers', University of Twente, (), pp. 2 [Online]. Available
at: http://essay.utwente.n l/71706/1/van%20Eeuwen_MA_BMS.pdf(Accessed: 20/11/2017)
Teller Vision. (2017). 'HGS Identifies Top 10 Customer Experience Trends', Issues and Trends, (), pp.
4 [Online]. Available at: http://eds.a.ebscohost.com/eds/pdfviewer/pdfviewer?vid=14&sid=1fd6a005-
b85b-430e-b3c8-30a3f70d947e%40sessionmgr4007 (Accessed: 21/11/2017).
Bayan Abu Shawar, B. A. and Eric Atwell, E. (2007). 'Chatbots: Are they Really Useful?', LDV-
Forum, pp. 29, 30
[Online].Available: http://eprints.whiterose.ac.uk/81930/1/AComparisonBetweenAliceElizabeth.pdf (
Ac
cessed:22/11/2017).
OLGA ANNENKO, O. (2016). When to Use Webhooks vs. APIs To Sync Data Between
Applications, Available at: https://www.elastic.io/sync-data-between-applications-apis-
webhooks/ (Accessed: 01/03/2018).
6
DialogFlow. (2015). ‘ A Speech Interface in 3 Steps’, Available

at: https://blog.dialogflow.com/post/speech-interface-3-steps/ (Accessed: 01/03/2018).
DZone. (20918). ‘A Writer's Guide to Conversational Interfaces’, Available
at: https://dzone.com/articles/a-writers-guide-to-conversational-interfaces?fromrel=true(Accessed:
20/03/2018).
Taylor Otwell, T. (2018). Laravel, Available at: https://laravel.com/docs/5.5/blade (Accessed:

05/02/2018).
cs.hmc.edu. (2018).’ Domain Specific Design Patterns: Designing For Conversational User
Interfaces’. [online] Trento: cs.hmc.edu. Available at:
https://arxiv.org/ftp/arxiv/papers/1802/1802.09055.pdf [Accessed 14 Mar. 2018].
Google. (2018). ‘Design Principles and Methodology’, Available
at: https://developers.google.com/actions/design/principles (Accessed: 27/03/2018).
http://www.spokenlanguagetechnology.com/docs/McTear_ESSV_2018.pdf.
(Accessed d04/03/2018)
Asbjørn Følstad, A. and, Petter Brandtzæg, P. (August 2017). ‘ Chatbots and the new world of
HCI’, Available
at: http://interactions.acm.org/archive/view/july-august-2017/chatbots-and-the-new-world-of-
hci (Accessed:
14/02/2018).
Open Banking Ltd 2018 (2018) Open Banking, Available

at: https://www.openbanking.org.uk/about-us/ (Accessed: 18/03/ 2018)
Facebook.(2018). ‘Messenger Platform Changes in Development’. Available
at: https://messenger.fb.com/newsroom/messenger-platform-changes-in-
development/ (Accessed:02/05/2018).
7
7.Appendices
The appendices are an opportunity to provide secondary material in support of the description in the
body of the report. In principle, the reader need not look at the appendices and no specific marks are
awarded for this section.
Sample content:
8
7.1 Appendix A Questionnaire Results User Requirements
9
10
11
Figure 7.1.1 responses to user that have used a chat bot before
12
13
Figure 7.1.2 results displayed for how secure users feel online banking is on Survey Monkey
14
15
16

2018 Thesis Dana Doherty - Chatbots

Uploaded by

Copyright:

Available Formats

2018 Thesis Dana Doherty - Chatbots

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2018 Thesis Dana Doherty - Chatbots

Uploaded by

Copyright:

Available Formats

L

EEE521 Final Year Project Report

BSc Hons Computer Science

Task-based Interaction Chatbot

Supervisor Dr. Kevin Curran

Machine Learning technology.

are being used to improve many services. In

This project aimed to implement a web-based chatbot to assist with online

is becoming more prevalent as technology companies are streamlining Artificial

1.1 Aims & Objective

The application will allow users to c:

• SMS confirmations and email mini statement schedule. The Starling

1.2 Literature Review

1.3 Chatbots in Industry

more efficient and immediate information to customers in comparison to a customer service

Figure 1.3.2: Industries piloting chatbots (Source: Etlinger,. , 2017)

1.4 Natural Language Understanding Engine

1.5 Artificial Intelligence

1.5.1 Artificial Intelligence Methods

1.6 Chatbot Architecture

1.7 Summary of Report

The remainder of this report will focus on:

2. Requirements Specification and Analysis

• the problem the project is overcoming

2.1 Problem Statement

2.2 Proposed solution

2.3 Software Development Methodologies

2.3.1 Waterfall Methodology

analysis, design, implementation and testing. Each stage is completed subsequently

2.3.2 Incremental Model

It is very common for software to be released in stages, it is critical that component

2.3.3 Chosen Methodology

2.4 Information Gathering Techniques

data of a large group quickly. Surveys consist of

Interviews : interviews are a structured approach to information gathering. They are

Discovery Prototyping: this involves developing a low fidelity throwaway prototype of

2.4.1 Chosen Information Gathering Technique

2.5 Functional and Non-Functional Requirements

availability, configurability and design constraints (Frank Tsui et al,.., 2016)

2.5.1 Functional Requirements

2.5.2 Non-Functional Requirements

2.6 Software and Hardware Specification

2.6.1 Software Requirements

2.6.2 Hardware Requirements

2.7 Project Planning

2.7.1 Project Milestones and Deliverables

Table 1: Project Milestones.

Milestone Milestone Description Completion Date

1 Literature Review 06/11/2017

2 Analysis & Requirements Specification 21/11/2017

3 Complete Interim Report 24/11/2017

8 Iteration 4 –final build 26/03/2018

Table 1: Project Milestones

Deliverable Deliverable Description

2 Web based Chatbot

Table 2: Project Deliverables

2.7.2 Development Iteration Breakdown

Iteration Functionality to be implemented

Iteration 1 • Install and configure VirtualBox VM and Vagrant