Probabilistic query expansion using query logs

View FullText article

• ACM, DOI

Muy buen ejemplo de correlación entre queries y documentos, query terms y document terms, usando ckickthrough data (logs), con formalización probabilística

Reviewed by sergiolopez as - 2008-11-15 16:40:35

First, we will test the assumption that the terms used in queries and in documents are truly very different. This assumption has often been made, but never tested by a quantitative measurement. Our test will show that there is indeed a large difference between the query terms and document terms

But for short queries, phrases are of crucial importance because they are more accurate representations of information and requirements. Without phrases, separate words in the query may lead to poor results. For example, given the query “search engine”, if it is represented as “search” and “engine”, few of the retrieved documents will be related to search engine, and most of them pertain to mechanical engines

The central idea of our method is that if a set of documents is often selected for the same queries, then the terms in these documents are strongly related to the terms of the queries. Thus some probabilistic correlations between query terms and document terms can be established based on the query logs.

• sergiolopez, aliku, JeffreyPalmer, dkaz, suripi, brusilovsky, TessaC, vitalaswp4, pprett, social_navigation (group), avirr, dreamedge

• clickthrough-data, eit, expansion, feedback, ir, log, logmining, log-mining, logs, mining, no-tag, probabilistic-retrieval, probability, query, query-expansion, querylog, query-log, query-optimization, quey, relevance, search-ir-eval, search-log-analysis, search-logs, social-search, usage, web, www2002, www-search