非表示:
キーワード:
-
要旨:
The ongoing explosion of web information calls for more intelligent and
personalied methods towards better search result quality for advanced queries.
Query log and click streams obtained from web browsers or search engines can
contribute to better quality by exploiting the collaborative recommendations
that are implicitly embedded in this information. This paper presents a new
method that incorporates the notion of query nodes into PageRank model and
integrates the implicite relevance feedback given by click streams into the
automated process of authority analysis. This approach generalizes the
well-known random-surfer model into a random-expert model that mimics the
behavior of an expert user in an extended session consisting of queries, query
refinements, and result-navigation steps. The enhanced PageRank scores, coined
QRank scores, can be computed offline; at query-time they are combined with
query-specific relevance measures with virtually no overhead. Our preliminary
experiments, based on real-life query-log and click-stream traces from eight
different trial users indicate significant improvements in the precision of
search results.
