A Framework for User Guidance in Web Search Engine Interfaces Based on Past Users' Behaviour

In this thesis, an adaptive Web search engine model is developed that assists its users in preparing relevant queries by recommending the related frequent phrases mined from previously submitted queries. The model also reorders the recommended pages of the conventional Web search engines based on the users' interests. Search engine query log mining has evolved over time to data stream mining due to the endless and continuous sequence of queries known as a query stream. We propose an Online Frequent Sequence Discovery (OFSD) algorithm to extract frequent phrases from within query streams based on a new frequency rate metric which is suitable for query stream mining. OFSD is an online, single pass, and real-time frequent sequence miner appropriate for data streams. The frequent phrases extracted by the OFSD algorithm are used to guide novice users to complete their search queries more efficiently. A re-rank method for the retrieved pages of a conventional Web search engine is also proposed which relies on past users clicks' for each frequent phrase extracted by OFSD. The contribution of our proposed model is three-fold. First, a Complementary Phrase Recommender module suggests a list of complementary phrases that are syntactically compatible with the entered query segment. Second, a Semantic Phrase Adviser module provides a list of the phrases that are semantically related to the entered query segment. These two modules help the user enter the most related phrases to his/her intention as a query. Third, a Page Rank Reviser module refines the order of the recommended documents prepared by a conventional Web search engine to help the user find the related Web pages at the top of the list. Two query logs with different characteristics are used to evaluate the proposed model. The experimental results confirm the significant benefit of monitoring frequent phrases within the queries instead of using the whole query as a non-separable item. The number of the monitored elements substantially decreases, which results in smaller memory consumption as well as better performance. YourEye, our implemented adaptive Web search engine based on the proposed model adjusted for the University of New Brunswick is introduced. Evaluation of YourEye by real users confirms the efficiency of the proposed model in performance as well as user satisfaction.