Embed Size (px)
Citation preview
A Generalized Framework of Exploring
Category Information for Question Retrieval in
Community Question Answer Archives
Xin Cao, Gao Cong, Bin Cui, and Christian S. Jensen
In Proceedings of the 19th international conference on world wide web
(WWW 2010)
Prepared and Presented by Baichuan Li
Outline
Introduction & Motivation
Category-Enhanced Question Retrieval
Models
Experiments
Conclusion
14/9/2010 Paper Presentation 2/21
Introduction
Community Question-Answering (CQA)
Services
14/9/2010 3/21 Paper Presentation
Question Retrieval
14/9/2010
Query
Existed similar
questions and
their answers
4/21 Paper Presentation
Motivation
14/9/2010 5/21
Query Category
Paper Presentation
CATEGORY-ENHANCED QUESTION RETRIEVAL MODELS
14/9/2010 Paper Presentation
Exploiting Categories in Question
Retrieval Given a query q, a historical question d,
and the category cat(d) that contains d:
where Sq,d is the local relevance score
and Sq,cat(d) is the global relevance score, N()
is the normalization function and α is a
weighting parameter.
Words play different roles in computing
local and global relevance scores
14/9/2010 7/21 Paper Presentation
Retrieval Models
Vector Space Model
Okapi BM25 Model
Language Model
Translation Model
Translation-Based Language Model
14/9/2010 Paper Presentation 8/21
Vector Space Model
14/9/2010 Paper Presentation 9/21
Vector Space Model
14/9/2010
Global relevance score
Local relevance score
Paper Presentation 10/21
Okapi BM25 Model
14/9/2010 Paper Presentation 11/21
Okapi BM25 Model
14/9/2010
Global relevance score
Local relevance score
Paper Presentation 12/21
Language Model
14/9/2010 Paper Presentation 13/21
Language Model
14/9/2010
Global relevance score
Local relevance score
d -> Cat(d)
Coll -> Cat(d)
Paper Presentation 14/21
Translation Model
14/9/2010
IBM translation models: http://en.wikipedia.org/wiki/Statistical_machine_translation
Paper Presentation 15/21
Translation Model
14/9/2010
Global relevance score
Local relevance score
d -> Cat(d)
Coll -> Cat(d)
Paper Presentation 16/21
Translation-Based Language Model
14/9/2010
Global relevance score
Local relevance score
d -> Cat(d)
Coll -> Cat(d)
Paper Presentation 17/21
EXPERIMENTS
14/9/2010 Paper Presentation
Data Set
Question Repository
Query Set
◦ 252 queries from http://homepages.inf.ed.ac.uk/gcong/qa
14/9/2010 Paper Presentation 19/21
Results
14/9/2010 Paper Presentation 20/21
Conclusion
Exploiting category information
associated with questions for improving
question retrieval
Conducting experiments with large scale
CQA data
Improvements
◦ Considering answers
◦ Utilizing hierarchical category structures
◦ …
14/9/2010 21/21 Paper Presentation
