Web Text Classification

1
Web Text Classification
2
Outline of Presentation

• Background Study
• Literature Review
• Evaluation Metrics
• References

3
Background StudyText Classification

• Text Classification
• A sub-field of information retrieval and text
  mining.
• Text documents are automatically organized into
  different pre-specified categories to facilitate
  document retrieval and subsequent analysis.
• Text documents are represented in term frequency
  matrix based on text indexing methods.
• Use soft-computing techniques, such as
• Naïve Bayesian Classifier
• Support Vector Machine

4
Background StudyText Classification Cont

• Limitations of representing documents using term
  features
• Polysemy
• One word can carry different meaning.
• Synonymy
• Several words carry same meaning.
• Local or global context
• The main concept available in whole content, one
  whole paragraph or one full sentence.

5
Background StudyInformation Extraction (IE)

• The task of locating relevant and specific
  information from text using statistical and
  linguistic analysis (semantic and syntactic
  analysis).
• Key element a set of text extraction rules or
  natural language processing (NLP) patterns that
  identify specific information to be extracted.
• Knowledge engineering approach
• Rules are hand-crafted by domain experts.
• Automatic training approach
• Run training algorithm on a set of annotated
  training corpus.
• Instead of a set of keywords, documents can be
  represented based on the NLP patterns and
  extracted information patterns feature.

6
Background StudyAssociation Rule Mining (ARM)

• Identify interesting relationships among items in
  a given data set.
• Let I i1, i2, , im be a set of m items. Let
  D, be as set of database transactions where each
  transaction T is a set of items such that T ? I.
• An association rule is an implication of the form
  A ? B, where A ? I, B ? I, and A ? B ?.
• Example of association rule generated from a
  transactional database
• vanilla_coke ? ice-cream
• support 5, confidence 83

7
Outline of Presentation

• Background Study
• Literature Review
• Evaluation Metrics
• References

8
Literature ReviewText Classification

• Web-page Classification through Summarization
  Shen et al., 2004
• Locate the content body of the Web pages as the
  basic summarisation component.
• Full-text summarization is performed on the page
  content.
• The summaries are then passed to standard text
  classification algorithms, such as Naïve Bayesian
  and Support Vector Machine.
• Average precision achieved is 80 81.

9
Literature ReviewIE for Text Classification

• Using IE to Classify Newspapers Advertisements
  Peleato et al., 2000
• All advertisements are classified into different
  categories using Naïve Bayesian classifier.
• Extract fillers for the templates of each
  category respectively, where each slot carries
  different weights (manually defined). A score is
  then given for each filled template.
• Results of both modules are compared, an
  advertisement is considered as unclassified if
  different results are produced by two methods.

10
Literature ReviewNLP for Text Classification

• Ntoulas et al., 2005 applied linguistic
  analysis in searching and ranking Web pages.
• Linguistic analysis operations POS tagging,
  phrase identification and named entity
  recognition.
• Classification of results are performed to solve
  word sense disambiguation.
• No full sentence parsing, the nature of concepts
  extracted is different from our proposal.

11
Literature ReviewIE for Text Classification

• IE as a Basis for High-Precision Text
  Classification Riloff and Lehnert, 1994
• Documents are represented in NLP
  annotations/patterns and extracted attributes
  values.
• Each NLP patterns is associated to a category.
• A document is classified as long as there is one
  single relevant NLP pattern in the document.

12
Outline of Presentation

• Background Study
• Literature Review
• Evaluation Metrics
• References

13
Evaluation Metrics

• Evaluating the classification results
• Precision
• Recall
• F-measure

14
Outline of Presentation

• Background Study
• Literature Review
• Evaluation Metrics
• References

15
References

• D. Shen, Z. Chen, Q. Yang, H.J. Zeng, B. Zhang,
  Y. Lu and W.Y. Ma, Web-page Classification
  through Summarization, Proceedings of the 27th
  annual International ACM SIGIR Conference on
  Research and Development in Information
  Retrieval, Sheffield, United Kingdom, 2004, pp
  242 246.
• A. Ntoulas, G. Chao and J. Cho, The Infocious
  Web Search Engine Improving Web Searching
  Through Linguistic Analysis, Proceeding of the
  14th International Conference on World Wide Web,
  May 2005, Chiba, Japan, ACM Press, pp 840 849.
• M.L. Antonie and O.R. Zaiane, Text Document
  Categorization by Term Association, Proceeding
  of IEEE International Conference on Data Mining,
  2002, pp 19 26.
• R.A. Peleato, J.C. Chappelier and M. Rajman,
  Using Information Extraction to Classify
  Newspapers Advertisements, Proceedings of the
  5th International Conference on the Statistical
  Analysis of Textual Data, Lausanne, Switzerland,
  March 2000.

16
References

• B. Liu, W. Hsu and Y. Ma, Integrating
  Classification and Association Rule Mining,
  Proceeding of The 4th International Conference on
  Knowledge Discovery and Data Mining, New York,
  1998, AAAI Publication, pp 80 86.
• B. Liu, Y. Ma and C.K. Wong, Classifications
  using Association Rules Weakness and
  Enhancements, in Vipin Kumar, et al, (eds), Data
  mining for Scientific Applications, 2001.
• A.A. Freistas, Understanding the Crucial
  Differences Between Classification and Discovery
  of Association Rules A Position Paper, ACM
  SIGKDD Explorations, July 2000, Vol. 2, Issue1,
  pp 65 69.
• L. Eikvil, Information Extraction from World
  Wide Web A Survey, published by Norwegian
  Computing Centre, Norway, July 1999.
• E. Riloff and W. Lehnert, Information Extraction
  as a Basis for High-Precision Text
  Classification, in ACM Transactions on
  Information Systems, Vol. 12, No. 3, July 1994,
  pp 296 - 333.

17
References

• D. Barbará, C. Domeniconi, N. Kang, Mining
  Relevant Text from Unlabelled Documents. In
  Proceedings of the Third IEEE International
  Conference on Data Mining (ICDM 03), 2003, pp
  489 492.
• E. Riloff and W. Phillips, An Introduction to
  the Sundance and Autoslog Systems, Technical
  Report, School of Computing, University of Utah,
  November 2004, USA.
• R. Kosala and H. Blockeel, Web Mining Research
  A Survey, ACM SIGKDD Explorations, July 2000,
  Vol. 2, Issue1, pp 1 15.
• X.Y. Gao, M.J. Zhang and P. Andreae, Automatic
  Pattern Construction For Web Information
  Extraction, in International Journal of
  Uncertainty, Fuzziness and Knowledge-Based
  Systems, Vol. 12, No. 4, 2004, pp 447 470.
• R.J. Mooney and R. Bunescu, Mining Knowledge
  from Text Using Information Extraction, ACM
  SIGKDD Explorations, June 2005, Vol. 7, Issue1,
  pp 3 10.
• J. Han and M. Kamber, Data Mining Concepts and
  Techniques, Morgan Kauffman Publishers, United
  States of America, 2001.