Facet-based opinion retrieval from blogs
|Facet-based opinion retrieval from blogs|
|Citation:||Information Processing and Management 46 (1): 71-88. 2010|
|Database(s):||Google Scholar cites|
|Web:||Bing Google Yahoo! — Google PDF|
|Article:||BASE Google Scholar PubMed|
|Restricted:||DTU Digital Library|
|Extract:||Talairach coordinates from linked PDF: CSV-formated wiki-formated|
- Blogs06, Blog07, Blog08 from TREC
- Sentiment word resources ("consisted of 6553 lexical units") from:
 Query expansion
The method for identifying "concepts" (a keyword/keyphrase/named entity) using Wikipedia was similar to the method described in Improving complex interactive question answering with Wikipedia anchor text. The query (first the entire part then subphrases) is matched with Wikipedia article titles. Query expansion is based on the redirects of Wikipedia. An example is given in Table 3.
 Opinion-based re-ranking
"opinion-based re-ranking consists of adjusting the tf weights of query terms on the basis of their cooccurrence with subjective lexical units in fixed-size windows centered on the query term occurrences."
Results for query expansion for the 3 blog corpus are given in Table 2.