Bonfring International Journal of Software Engineering and Soft Computing, Vol. 9, No. 2, April 2019
47
Effective Online Discussion Data for Teacher’s
Reflective Thinking Using Feature base Model
V. Yasvanth Kumaar and Dr.G. Singaravel
Abstract--- In this paper analysis automatic coding method
by integrating the inductive content analysis and text
classification techniques. In existing model acquire the
reflective thinking categories by conducting an inductive
content analysis and base our text classification algorithm on
the categories, so we augment the manual method of coding.
We apply the trained classification model to a large-scale and
unexplored online discussion data set, so we can have a
comprehensive understanding of teachers’ reflection. This
paper also provides six types of visualizations of the text
classification results: the visualization of teachers’ reflection
level and the visualization of teachers’ reflection evolution. By
using the categories gained from inductive content analysis to
create a radar map, we visually represent teachers’ reflection
level after obtain the results of text classification future
classification model.
Keywords--- Data Mining, Teacher Reflection Model, TFIDF Classification, Visualization Learning.
I.
INTRODUCTION
D
ATA mining is that the method of extracting patterns
from knowledge. Data mining is seen as a progressively
vital tool by fashionable business to transform knowledge into
an informational advantage. It’s presently utilized in a wide
range of identification practices, like selling, surveillance,
fraud detection, and scientific discovery.
The connected terms knowledge dredging, knowledge
fishing and knowledge snooping confer with the utilization of
knowledge mining techniques to sample parts of the larger
population data set that are (or might be) too tiny for reliable
statistical inferences to be created regarding the validity of any
patterns discovered (see additionally data-snooping bias).
These techniques will but, be utilized in the creation of latest
hypothesizes to check against the larger knowledge
populations.
Data mining is that the method of applying these strategies
to knowledge with the intention of uncovering hidden patterns.
It’s been used for several years by businesses, scientists and
governments to sift through volumes of information like
airline passenger trip records, census knowledge and food
market scanner knowledge to supply research reports. (Note,
V. Yasvanth Kumaar, PG Scholar, Department of Information
Technology, K.S.R. College of Engineering, Tiruchengode, India.
E-mail: yasvanth123456@gmail.com
Dr.G. Singaravel, Professor & Head, Department of Information
Technology, K.S.R. College of Engineering, Tiruchengode, India.
E-mail: singaravelg@gmail.com
DOI:10.9756/BIJSESC.9022
however, that coverage isn't perpetually thought-about to be
data processing.)
In this study, we encoded and visualized the large-scale
unstructured text data in teachers’ online collaborative
learning activities in order to understanding teachers’
reflection levels and evolution, integrating both the qualitative
content analysis method and educational data mining
techniques. The qualitative content analysis method revealed
that teachers’ reflective thinking included TechnicalDescription,
Technical-Analysis,
Technical-Critique,
Personalistic-Description,
Personalistic-Analysis,
and
Personalistic-Critique. Based on the results of inductive
content analysis, we implemented a single-label text
classification algorithm to classify the sample data. Then we
applied the trained classification model on a large-scale and
unexplored online discussion text data set. After the online
discussion text data being classified, two types of
visualizations of the results were provided. By using the
categories gained from inductive content analysis to create a
radar map, teachers’ reflection level had been represented. In
addition, by defining the variables of the coding scheme as the
nodes and the co-occurrences of nodes within a discussion
post as the connections, a cumulative adjacency matrix was
created to characterize the evolution of teachers’ reflective
thinking.
II.
RELATED WORKS
Weize Kong and James Allan [1] evaluated Faceted Web
Search systems by their utility in assisting users to clarify
search intent and subtopic information. The authors described
how to build reusable test collections for such tasks, and
propose an evaluation method that considers both gain and
cost for users. Faceted search enables users to navigate a
multi-faceted information space by combining text search with
drill-down options in each facet. For example, when searching
\computer monitor" in an e-commerce site, users can select
brands and monitor types from the the provided facets:
fSamsung, Dell, Acer,... g and f LET-Lit, LCD, OLEDg. This
technique has been used successfully for many vertical
applications, including e-commerce and digital libraries.
Krisztian Balog et al., [2] consider the task of entity
search and examine to which extent state-of-art information
retrieval (IR) and semantic web (SW) technologies are capable
of answering information needs that focus on entities. We also
explore the potential of combining IR with SW technologies to
improve the end-to-end performance on a specific entity
search task. We arrive at and motivate a proposal to combine
text-based entity models with semantic information from the
Linked Open Data cloud. The problem of entity search has
ISSN 2277-5099 | © 2019 Bonfring
Bonfring International Journal of Software Engineering and Soft Computing, Vol. 9, No. 2, April 2019
been and is being looked at by both the Information Retrieval
(IR) and Semantic Web (SW) communities and is, in fact,
ranked high on the research agendas of the two communities.
The entity search task comes in several flavors. One is known
as entity ranking (given a query and target category, return a
ranked list of relevant entities), another is list completion
(given a query and example entities, return similar entities),
and a third is related entity finding (given a source entity, a
relation and a target type, identify target entities that enjoy the
specified relation with the source entity and that satisfy the
target type constraint.
Chengkai Li, Ning Yan et al [3] focused on automatic
and dynamic faceted interfaces. The facets could not be precomputed due to the query-dependent nature of the system. In
applications where faceted interfaces are deployed for
relational tuples or schema-available objects, the
tuples/objects are captured by prescribed schemata with
clearly defined dimensions (attributes), therefore a queryindependent static faceted interface (either manually or
automatically generated) may suffice. By contrast, the articles
in Wikipedia are lacking such predetermined dimensions that
could fit all possible dynamic query results. Therefore efforts
on static facets would be futile.
Wisam Dakka et al., [4] presented a set of techniques for
automatically identifying terms that are useful for building
faceted hierarchies. The techniques build on the idea that
external resources, when queried with the appropriate terms,
provide useful context that is valuable for locating the facets
that appear in a database of text documents. It is demonstrated
the usefulness of Wikipedia, WordNet, and Google as external
resources. Experimental results, validated by an extensive
study using human subjects, indicate that our techniques
generate facets of high quality that can improve the browsing
experience for users. If efficiency is not a major concern, can
incorporate multiple such resources in this framework, for a
variety of topics, and use all of them, irrespectively of the
topics that appear in the underlying collection. The
distributional analysis step of our technique automatically
identifies which concepts are important for the underlying
database and generates the appropriate facet terms.
Amaç Herdagdelen et al., [5] presents a novel approach
to query reformulation which combines syntactic and semantic
information by means of generalized Levenshtein distance
algorithms where the substitution operation costs are based on
probabilistic term rewrite functions. We investigate
unsupervised, compact and efficient models, and provide
empirical evidence of their effectiveness. Further it explores a
generative model of query reformulation and supervised
combination methods providing improved performance at
variable computational costs. Among other desirable
properties, our similarity measures incorporate informationtheoretic interpretations of taxonomic relations such as
specification and generalization.
X. Xue and W. B. Croft [6] propose a novel framework
where the original query is transformed into a distribution of
reformulated queries. A reformulated query is generated by
applying different operations including adding or replacing
query words, detecting phrase structures, and so on. Since the
48
reformulated query that involves a particular choice of words
and phrases is explicitly modeled, this framework captures
dependencies between those query components. On the other
hand, this framework naturally combines query segmentation,
query substitution and other possible reformulation operations,
where all these operations are considered as methods for
generating reformulated queries. In other words, a
reformulated query is the output of applying single or multiple
reformulation operations.
Mariana Damova, Ivan Koychev [8] presents a survey of
recent extractive query-based summarization techniques. We
explore approaches for single document and multi-document
summarization. Knowledge-based and machine learning
methods for choosing the most relevant sentences from
documents with respect to a given query are considered.
Further, expose tailored summarization techniques for
particular domains like medical texts. The most recent
developments in the fled are presented with opinion
summarization of blog entries. This survey is motivated by the
idea of making e-books more intelligent, in particular enabling
them to “answer” users’ queries. To find the needed
information in books users usually do not want to spend a long
time searching, browsing or skimming them. They will be
happy to have a “guru” nearby that can provide them with the
right answer almost simultaneously. For this purpose to had a
close look at the area of automated text summarization.
Recently, with the increasing of information available online,
those approaches have been developed very extensively. In the
realm of automatic summarization different kinds of
summarization have been attempted. Along with the study
distinguish between the following types of summaries
according to specific criteria.
III.
METHODOLOGY
The purposes of this research are 1) to explore the
categories of teachers’ reflective thinking and realize the
automatic classification of the online discussion data by
integrating the inductive content analysis and educational data
mining techniques; and 2) to analyze teachers’ reflective
thinking in the online teacher professional development
program, including teachers’ reflection levels and evolution.
Teachers’ online discussion data has been collected and
analyzed for understanding their reflective thinking mainly
because:
•
•
•
•
The peer coaching model has long been used in
professional development programs to enhance
teachers’ teaching practices and students’ learning
out-comes. In reciprocal peer coaching, teachers.
Share the coaching role, discuss with each other, and
elicit the reflection [10]. Teachers’ online discussion
data embodies their reflective thinking.
Based on the understanding of teachers’ reflective
thinking, teacher training managers and educational
researchers can design more appropriate online
learning activities and provide proper services and
interventions to support teachers’ online reflection.
Content analysis is commonly used to analyze
transcript of online discussion for educational
ISSN 2277-5099 | © 2019 Bonfring
Bonfring International Journal of Software Engineering and Soft Computing, Vol. 9, No. 2, April 2019
purposes. In this study, the coding frameworks, which
focused on the main content and level of teachers’
reflection, had been obtained from inductive content
analysis. In addition, we had developed a Data
collector, the Hawk, to collect posts on OPDP.
Therefore, we chose to analyze teachers’ posts for
understanding their reflective thinking
Stem Word Document
In this module, enter the given word and stem word using
text box control and click save button stem word saved into
the table. The details are saved in ‘Stemword’ file. The stem
word details view on view controls.
specific technical terms), so they are diverse and have various
synonyms. For example, in the sentence “Trust influences
communication between organizations”. To identify the Trust
and communication is an entity of success factor and
influences as an influencing relationship.
The relevant and irrelevant sentences are annotated into
positive class and negative class respectively. The annotation
is manually performed by a domain expert based on the given
related keywords and those keywords are used as a guideline
for the annotation. In this module, user can submit the certain
keywords which are stored into the KeyWords file.
Subset of Six Levels Preparation
Stop Word Document
In this module, enter the stop word using text box control
and click save button stop word saved into the table. The
details are saved in ‘Stopword’ file. The stop word details
view on view controls.
Add Synonym Word Document
In this module, enter the given word and synonym word
using text box control and click save button synonym word
saved into the table. The details are saved in ‘synonym word’
file. The synonym word details view on view controls.
Add Test Dataset
The test dataset in this context are different which
technical terms are, and hence they are fixed-terms. In other
words, the keywords are natural language like (i.e., not
Teacher Generate Online
Discussion Data
•
•
•
•
Data Storage
1
49
Phase 1: A data collection tool to collect teachergenerated online discussion data and obtained 21,388
posts by using the data collection tool over a period
of 6 months.
Phase 2: A random sample (2,000 posts) were drawn
from the online discussion data set and used for
inductive content analysis.
Phase 3: Three experts who had experience with
qualitative research and familiarity with pre-existing
coding schemes collaborated on the inductive content
analysis process. Three experts conducted an
inductive content analysis on the random sample of
online discussion data set.
After phase 3, we obtained the categories used for
text classification.
2
Data collect Users/ Research
6
3
Classification Model
(TF-IDF)
Large Scale Text Classified
4
Code Scheme
5
7
Update Test
Teacher Generate Online
Discussion Data
Classification Model
(Conational-TF-IDF)
Data Storage
9
8
ISSN 2277-5099 | © 2019 Bonfring
Bonfring International Journal of Software Engineering and Soft Computing, Vol. 9, No. 2, April 2019
•
•
•
Phase 4: A single-label NaıveBayes Classification
algorithm had been implemented to classify the
labeled data. Then, we evaluated the performance of
the single-label NaıveBayes classification algorithm
by comparing it with other commonly used text
classification algorithm.
Phase 5: Implemented a large-scale text classification
based on the trained classification model. All the
online discussion data has been automatically coded.
Phase 6: Visually represented teachers’ reflection
levels after we obtained the results of large-scale text
classification. Additionally, we created a cumulative
adjacency matrix to characterize the evolution of
teachers’ reflective thinking. Then, the results of text
classification and visualization were sent back to the
data storage.
Online Collaborative Learning Approach
The approach of online collaborative learning provides
Internet-based professional development opportunities,
including individual reflection, sharing resource, work- shops,
online interactions with colleagues, and mentors
•
•
•
•
•
In the first stage, every teacher reflected the
discussion topic individually. After the chief teacher
posted the activity plan and discussion topic onto the
OPDP, every teacher downloaded and read the plan
and topic
In the second stage, all teaches discussed the topic
collectively. Every teacher expressed their own
views, and exchanged views with others. This is a
process of externalization. Teachers expressed their
ideas or thoughts about a specific topic via posts.
In the final stage, every teacher submitted a
document, which recorded his/her learning
experience in the online collaborative learning
activity. This is a process of internalization. Teachers
recorded the knowledge learned from communication
with mentors and colleagues.
Each three-stage online collaborative learning
activity usually lasted for one month. After three
stages were completed, the chief teacher and the
individual teachers summarized the results.
In the three-stage online collaborative learning
activity, teachers’ reflection thinking was expressed
in two main ways: the online discussion posts and the
reflection documents. In this study, we collected
teachers’ online discussion posts to analyze their
reflective thinking.
Term Frequency – Inverse Document Frequency (TF-IDF)
The TF measures how frequently a particular term occurs
in a document. It is calculated by the number of times a word
appears in a document divided by the total number of words in
that document. It is computed as TF (the) = (Number of
times term the ‘the’ appears in a document) / (Total
number of terms in the document). The IDF measures the
importance of a term. It is calculated by the number of
documents in the text database divided by the number of
documents where a specific term appears. While computing
50
TF, all the terms are considered equally important. That
means, TF counts the term frequency for normal words like
“is”, “a”, “what”, etc. Thus we need to know the frequent
terms while scaling up the rare ones, by computing the
following: IDF (the) = log_e(Total number of documents /
Number of documents with term ‘the’ in it).
For example, Consider a document containing 1000 words,
wherein the word give appears 50 times. The TF for give is
then (50 / 1000) = 0.05. Now, assume that, 10 million
documents and the word give appears in 1000 of these. Then,
the IDF is calculated as log(10,000,000 / 1,000) = 4. The TFIDF weight is the product
IV.
CONCLUSION
An inductive content analysis on samples taken from 17,624
posts was implemented and the categories of teachers’
reflective thinking were obtained. Based on the results of
inductive content analysis, we implemented a single-label text
classification algorithm to classify the sample data. Then, we
applied the trained classification model on a large-scale and
unexplored online discussion text data set and two types of
visualizations of the results were provided. By using the
categories gained from inductive content analysis to create a
radar map, teachers’ reflection level was represented. In
addition, a cumulative adjacency matrix was created to
characterize the evolution of teachers’ reflective thinking. This
study could partly explain how teachers reflected in online
professional learning environments and brought awareness to
educational policy makers, teacher training managers, and
education researchers. The considerably better results are
found for the planned technique compared to existing
strategies, regardless of the classifiers used. All the results
according in this paper demonstrate the practicability and
effectiveness of the planned technique. It’s capable of
distinctive co-regulated clusters of genes whose average
expression is strongly related to the sample classes. The
known gene clusters could contribute to revealing underlying
category structures, providing a useful gizmo for the
explorative analysis of biological information.
In future, the here fast algorithm enforced in information
set solely, additional planned to implement in numerical
information set. During this application, enforced cancer
dataset solely, in future planned to real time dataset like
diabetics, pressure so on. Additional a lot of, planned to
compare with algorithm like naïve bayes, K-Nearest neighbor
so on.
REFERENCES
[1]
[2]
[3]
[4]
W. Kong and J. Allan, “Extending faceted search to the general web”, in
Proc. ACM Int. Conf. Inf. Knowl. Manage., Pp. 839–848, 2014.
K. Balog, E. Meij and M. De Rijke, “Entity search: Building bridges
between two worlds”, In Proc. 3rd Int. Semantic Search Workshop, Pp.
9:1–9:5, 2010.
C. Li, N. Yan, S.B. Roy, L. Lisham and G. Das, “Facetedpedia:
Dynamic generation of query-dependent faceted interfaces for
Wikipedia”, in Proc. 19th Int. Conf. World Wide Web, Pp. 651–660,
2010.
W. Dakka and P.G. Ipeirotis, “Automatic extraction of useful facet
hierarchies from text databases”, In Proc. IEEE 24th Int. Conf. Data
Eng., Pp. 466–475, 2008.
ISSN 2277-5099 | © 2019 Bonfring
Bonfring International Journal of Software Engineering and Soft Computing, Vol. 9, No. 2, April 2019
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
A. Herdagdelen, M. Ciaramita, D. Mahler, M. Holmqvist, K. Hall, S.
Riezler and E. Alfonseca, “Generalized syntactic and semantic models
of query reformulation”, in Proc. 33rd Int. ACM SIGIR Conf. Res.
Develop. Inf. retrieval, Pp. 283–290, 2010.
X. Xue and W. B. Croft, “Modeling reformulation using query
distributions”, ACM Trans. Inf. Syst., Vol. 31, No. 2, Pp. 6:1–6:34,
2013.
L. Bing, W. Lam, T.L. Wong and S. Jameel, “Web query reformulation
via joint modeling of latent topic dependency and term context”, ACM
Trans. Inf. Syst., Vol. 33, No. 2, Pp. 6:1–6:38, 2015.
S.A. Gionis and Y. Maarek, “Improving recommendation for long-tail
queries via templates”, in Proc. 20th Int. Conf. World Wide Web, Pp.
47–56, 2011.
M. Damova and I. Koychev, “Query-based summarization: A survey”,
In Proc. S3T, Pp. 142–146, 2010.
L.K.R. Veni and R. Rajaram, “Afgf: An automatic facet generation
framework for document retrieval”, In Proc. Int. Conf. Adv. Comput.
Eng., Pp. 110–114, 2010.
J. Pound, S. Paparizos and P. Tsaparas, “Facet discovery for structured
web search: A query-log mining approach”, in Proc. ACM SIGMOD Int.
Conf. Manage. Data, Pp. 169–180, 2011.
W. Kong and J. Allan, “Extracting query facets from search results”, In
Proc. 36th Int. ACM SIGIR Conf. Res. Develop. Inf. Retrieval, Pp. 93–
102, 2013.
Y. Liu, R. Song, M. Zhang, Z. Dou, T. Yamamoto, M. P. Kato, H.
Ohshima and K. Zhou, “Overview of the NTCIR-11 imine task”, In
Proc. NTCIR-11, Pp. 8–23, 2014.
ISSN 2277-5099 | © 2019 Bonfring
51