User feedback session with clicked and unclicked documents for related search recommendation

Keyword based search is extensively used method to discover knowledge on the web. Generally, web users unable to arrange and define input queries relevant to their search because of adequate knowledge about domain. Hence, the input queries are normally short and ambiguous. Query recommendation is a method to recommend web queries that are related to the user initial query which helps them to locate their required information more precisely. It also helps the search engine to return appropriate answers and meet their needs. Usually users have ambiguous keywords in their mind to represent their information need. Hence, it is not a good idea to generate relation between user query keywords for recommendations. In this paper, we have presented Related Search Recommendation (RSR) framework, which discovers keywords which are present in snippets clicked and unclicked documents in feedback session. Pseudo documents are generated from feedback sessions which reflect what users wish to retrieve. Finally, semantic similarity is calculated between the terms present in pseudo document and used for recommendations. The proposed method provides semantically related search queries for the given input query. Simulation results show that the proposed framework RSR outperforms Rocchio's model and Snippet Click Model.