01 Intro

9/23/2020
Text Technologies for Data Science

INFR11145
Introduction
Instructor:
Walid Magdy
23-Sep-2020
Lecture Objectives
• Know about the course:
• Topic
• Objectives
• Requirements
• Format
• Logistics
• Note:
• No much technical content today
• Don’t assume next lectures would be the same!
2
Walid Magdy, TTDS 2020/2021
1
9/23/2020
Text Technologies for Data Science
= documents, words, terms, …

≠ images, videos, music (with no text)
Information Retrieval Search Engines

Text Classification Technologies
Text Analytics
3
What is Information Retrieval (IR)?
IR is NOT just
Web search
4
2
9/23/2020
What is IR?
Speech - QA
5
What is IR?
Information
Filtering
Recommendation
Social search
6
3
9/23/2020
What is IR?
Library (book) search

1950’s
7
What is IR?
Legal search
8
4
9/23/2020
What is IR?
Cross-Language search
9
What is IR?
Content-based music search

10
10
5
9/23/2020
*Source: Matt Lease (IR Course at U Texas)
What is IR?
Advertising
Query suggestion
/ correction
Snippet selection
/ summarisation
Categorisation
(search verticals)
11
11
What is IR? Find?
IR ≠ Find
• Sequential
• Exact match
12
12
6
9/23/2020
What is IR?
• IR is finding material of an unstructured nature that
satisfies an information need from within large
collections
• Find → Task
• Unstructured → Nature
• Information need → Target
• Satisfies → Evaluation
13
13
Text classification
14
14
7
9/23/2020
Text classification
15
15
Text classification
16
16
8
9/23/2020
What is text classification?

• Text classification is the process of classifying
documents into predefined categories based on their
content.
- Input: Text (document, article, sentence)

- Task: Classify into one/multiple categories
- Categories:
- Binary: relevant/irrelevant, spam .. etc.
- Few: sports/politics/comedy/technology
- Hierarchical: patents
17
17
In this course, we will learn to

• How to build a search engine
• which search results to rank at the top
• how to do it fast and on a massive scale
• How to evaluate a search algorithm
• is system A really better than system B
• How to work with text
• two tweets talk about the same topic?
• handle misspellings, morphology, synonyms
• How to classify text
• into categories (sports, news, comedy, …)
• features to use
• evaluate classification quality
• Apply text analytics
• Find what makes a set of document different from others
18
18
9
9/23/2020
How this course is different from others?

• ANLP, FNLP
• Some text processing
• Text laws
• No NLP (word/phrase level vs document level)
• ML practical
• Text classification
• No ML (using off-the-shelf ML tool)
• It does not overlap with others on:

• Search engines
• IR methods/models
• IR evaluation
• Text analysis
19
19
Some terms you will learn about

• Inverted index
• Vector space model
• Retrieval models: TFIDF, BM25, LM
• Page rank
• Learning to rank (L2R)
• MAP, MRR, nDCG
• Mutual information, information gain, Chi-square
• SVMs: binary/multiclass classification, ranking, regression
20
20
10
9/23/2020
This Course is Highly Practical

• 70% of the mark is on practical work
• You will implement 50+% of what you learn
• By W5, you should have developed a basic working
Search Engine from scratch
• Practical Lab every week
• Two coursework, mostly coding
• A course group project to develop a full system
21
21
Pre-requests (1/3)
• Maths requirements:
• Linear algebra: vectors/matrices (addition, multiplication, inverse,
projections ... etc).
• Probability theory: Discrete and continuous univariate random variables.
Bayes rule. Expectation, variance. Univariate Gaussian distribution.
• Calculus: Functions of several variables. Partial differentiation. Multivariate
maxima and minima.
• Special functions: Log, Exp, Ln.
22
22
11
9/23/2020
Pre-requests (2/3)
• Programming requirements:
• Python or Perl
• Knowledge in regular expressions
• Shell commands (cat, sort, grep, uniq, sed, ...)
• Data structures and software engineering for course project.
23
23
Pre-requests (3/3)
• Team-work requirement:
• Final course project would be in groups of 5-6 students.
• Working in a team for the project is a requirement.
• No exceptions will be allowed!
24
24
12
9/23/2020
Skills to be gained !!!

• Working with large text collections
• Few shell commands
• Some Python programming
• Software engineering skills
• Build text classifier in few mins
• TEAM WORK
• Project management
• Time management
• Task assignment + system integration
25
25
Course Structure
• 20 Lectures:
• 2 lectures → Introduction (today)
• 14 lectures → IR (50% practical lectures)
• 4 lectures → Text Analytics/Classification
• 8-10 Labs:
• Practice what you learn
• No Tutorials
• Some self-reading
• Lots of system implementation
• Few online videos
26
26
13
9/23/2020
Course Instructors
Walid Magdy Steve Wilson

Reader Research Associate
(14 lectures) (5 lectures)
+ 1 guest lecture
27
27
Lecture Format
• 2 Lectures at a time
• Questions are allowed any time. Feel free to interrupt
• 5-10 mins break after L1
• Feel free to go out and come back
• Discuss 1st lecture with friends
• Questions on L1 are allowed before starting L2
• Mind teaser math problem (for fun)
• Some lectures are interactive. Please participate
• Some lectures will include demos (running code)
• 2 tutorial lectures about using tools
28
28
14
9/23/2020
Labs
• Online!
• How it will work?
• Relevant lab will be announced with each lecture on Wednesday
• You should implement lab directly after lecture
• Any issues → ask on Piazza (tag question by lab number)
• Produced output → Share on Piazza (publicly)
• Demonstrators → answer questions + validate your output
• DO NOT ask a question before checking if it was asked before
• Tuesdays → Optional Teams meetings for those still require support
• Live lab times: Tuesday, 10am, 12am, 6pm

• Demonstrators:
Marina Posti, Maysara Hammouda and Zheng Zhao
29
29
Assessments
• Coursework 1: 10%
The same as labs 1-3 → Build your first search engine
• Coursework 2: 20%
IR Evaluation, Text classification/analytics
• Group project: 40%
A full running search engine supported by text technologies
• Final Exam: 30%
30
30
15
9/23/2020
Group Project
• The largest weight: 40% of the total mark
• Teamwork → Group 5-6 (you select your group)
• Design a full end-to-end search engine that searches a large
collection of documents with many functionalities.
• Mark:
• 50% on project → the same for all team members
- How complete/effective/fast/nice is your search engine?
• 50% on individual contribution → different for each member
- How useful/much is your contribution? (Mark can be negative!)
• Project prize → a prize will be awarded to best project
31
31
Example
• A search engine that retrieves quotes of movies and TV shows.
• Collection size: 77 million movie quotes
• Search options
• Phrase search of quotes
• Movie info search
• Advanced search: movie title, actor/actress, years, keywords
• Query suggestion
• Classifying results by genre
• Demo: http://167.71.139.222/
32
32
16
9/23/2020
Timeline
• 2 Semesters (or one?)
Exam
Lectures Labs
Semester 1 Semester 2
W5 W11 W11
CW 1 & 2 Group Project

33
33
Logistics (1/2)
• Lectures:
• Live on 2 Wednesdays, 12.00-14.30 (some exception might occur)
• Recording will be available
• Handouts to be posted on the day of the lecture
• Course webpage:
• Link: http://www.inf.ed.ac.uk/teaching/courses/tts/
• Handouts, Labs, CW details
• Learn:
• Lecture recordings
• Deadlines
34
34
17
9/23/2020
Logistics (2/2)
• Pizza:
• All communication will be there
• Questions about lectures/labs/CW are there
• Feel free to answer each other questions
• Lab support will be mainly there
• Please share your lab answers there
• Join NOW: link
• Microsoft Teams:
• Live lab support will be there
• Join NOW: link
35
35
FAQ
• How the project would be managed? What if one
member does not work?
• I am not that solid in programming, should I take this
course?
• Can I audit the course?
• Anything else?
36
36
18
9/23/2020
Next Lecture
• Definitions of IR main concepts
(more introduction)
37
37
19

01 Intro

Uploaded by

Copyright:

Available Formats

01 Intro

Uploaded by

Document Information

Original Title

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

01 Intro

Uploaded by

Copyright:

Available Formats

9/23/2020

Text Technologies for Data Science

Text Technologies for Data Science

= documents, words, terms, …

Information Retrieval Search Engines

What is Information Retrieval (IR)?

Library (book) search

Content-based music search

*Source: Matt Lease (IR Course at U Texas)

What is IR? Find?

What is text classification?

- Input: Text (document, article, sentence)

In this course, we will learn to

How this course is different from others?

• It does not overlap with others on:

Some terms you will learn about

This Course is Highly Practical

Skills to be gained !!!

Walid Magdy Steve Wilson

• Live lab times: Tuesday, 10am, 12am, 6pm

• Project prize → a prize will be awarded to best project

CW 1 & 2 Group Project

You might also like