Abstract: The important aim of each information retrieval system is to make available the keywords and applies refined ranking algorithms to categories them. But unfortunately, users give short and ambiguous queries which sometimes are not sufficient to clearly identify its requirement. Due to which incorrect result is formed, sorted and offered to the user. This report tends to propose innovative approach to infer user search goals by analyzing user query logs. First method used to get completely different user search goals for a query by cluster the feedback sessions. Second, a novel approach used to get pseudo-documents to better represent the feedback sessions for clustering. In addition, the task tail clustering method is used to automatically group queries into tasks. A task is defined to be an atomic user data need, whereas a task trail represents all user activities among that specific task, like query reformulations, URL clicks. Finally, two methods are compared, shows the task trail method is better than a feedback session to find user search goals. Moreover, propose a system intend Log Likelihood Ratio method to suggest related queries to the user. The experimental results on a dataset of queries prove the excellent performance of task trail method compared with feedback session methods and confirm the importance of identification of task. The performance of the system is measured in terms of processing time, recall, accuracy and recognition rate. The recognition rate of feedback session is 74.42 % and of task trail is 89.92%.

Keywords: User search goals, Query log, Task trail, Feedback session, Query suggestion, Cosine similarity, and Clustering.

| DOI: 10.17148/IARJSET.2022.9406