TY - GEN
T1 - Blog opinion retrieval based on topic-opinion mixture model
AU - Jiang, Peng
AU - Zhang, Chunxia
AU - Yang, Qing
AU - Niu, Zhendong
PY - 2010
Y1 - 2010
N2 - Recently, as blog is becoming a popular medium to express opinions, blog opinion retrieval excites interest in the field of information retrieval. It helps to find and rank blogs by both topic relevance and opinion relevance. This paper presents our topic-opinion mixture model based approach to blog opinion retrieval in the TREC 2009 blog retrieval task. In our approach, we assume each topic has its own opinion relevance model. A topic-opinion mixture model is introduced to update original query model, and can be regarded as a mixture of topic relevance model and opinion relevance model. By pseudo-relevance feedback method, we can estimate these two models from topic relevance feedback documents and opinion relevance feedback documents respectively. Therefore our approach does not need any annotated data to train. In addition, the global representation model is used to represent an entire blog that contains a number of blog posts. Experimental results on TREC blogs08 collection show the effectiveness of our proposed approach.
AB - Recently, as blog is becoming a popular medium to express opinions, blog opinion retrieval excites interest in the field of information retrieval. It helps to find and rank blogs by both topic relevance and opinion relevance. This paper presents our topic-opinion mixture model based approach to blog opinion retrieval in the TREC 2009 blog retrieval task. In our approach, we assume each topic has its own opinion relevance model. A topic-opinion mixture model is introduced to update original query model, and can be regarded as a mixture of topic relevance model and opinion relevance model. By pseudo-relevance feedback method, we can estimate these two models from topic relevance feedback documents and opinion relevance feedback documents respectively. Therefore our approach does not need any annotated data to train. In addition, the global representation model is used to represent an entire blog that contains a number of blog posts. Experimental results on TREC blogs08 collection show the effectiveness of our proposed approach.
KW - Blog
KW - Opinion retrieval
KW - Rank
KW - Topic-opinion mixture model
UR - http://www.scopus.com/inward/record.url?scp=79956310359&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-13672-6_25
DO - 10.1007/978-3-642-13672-6_25
M3 - Conference contribution
AN - SCOPUS:79956310359
SN - 3642136710
SN - 9783642136719
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 249
EP - 260
BT - Advances in Knowledge Discovery and Data Mining - 14th Pacific-Asia Conference, PAKDD 2010, Proceedings
T2 - 14th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2010
Y2 - 21 June 2010 through 24 June 2010
ER -