A Hybrid Method for Query Recommendation in Recommender Systems

The rapid growth of information in the Internet and high informational overload has created an important challenge for users in accessing their needed information. Nowadays, query recommendation systems have become a major part of information retrieval systems. One of the applications of these recommendation systems is in medical sciences. Through applying personalization approach, these systems attempt to decrease the problem of informational overload in Web and to accelerate user's medical search.
In this applied and descriptive study, by using the lexical features and search results of queries, we tried to propose a method that helps users to access their desired information in a short time while maintaining the lexical relationship with the original query. The popular k-means algorithm was used to cluster queries. The implementation of the proposed method was done by using java programming language and in NetBeans IDE software.
According to the proposed method, the combined use of the lexical features and search results of queries leads to useful information for detecting similar queries. Since there is a possibility that a query contains multi-meaning words, using search results can be useful in identifying the user’s intent of a query.
Evaluation of the proposed model with the real search log of the Parsijoo search engine indicated the precision rate of 77.4% for this method that in comparison to other methods shows 10% improvement of precision.