Posted on

Personalization aims to improve user’s searching experience by
tailoring search results according to individual user’s interests.
Typically, search engines employ two-level ranking strategy.
Firstly, initial list of documents is prepared using a low-quality
ranking function that is less computationally expensive. Secondly,
initial list is re-ranked by machine learning algorithms which
involve expensive computation. The proposed approach explores
the second level of ranking strategy which exploits user
information. In this approach, queries and search-result clicks are
used to model the user interest profiles probabilistically. The
user’s history provides the prior probability that a user searches
for a topic which is independent of user query. The document
topical features are combined with user specific information to
determine whether a document satisfies user’s information need
or not. The probability of relevance of each retrieved document
for a query is computed by integrating user topic model and
document topic model. Thus, documents are re-ranked according
to the personalized score computed for each document. The
proposed approach has been implemented and evaluated using
real dataset similar to AOL search log dataset for personalization.
Empirical results along with the theoretical foundations of the
model confirm that the proposed approach shows promising


Leave a Reply hello everyone

This site uses Akismet to reduce spam. Learn how your comment data is processed.