Personalized Information Retrieval based on Time-Sensitive User Profile

Abstract : Recently, search engines have become the main source of information for many users and have been widely used in different fields. However, Information Retrieval Systems (IRS) face new challenges due to the growth and diversity of available data. An IRS analyses the query submitted by the user and explores collections of data with unstructured or semi-structured nature (e.g. text, image, video, Web page etc.) in order to deliver items that best match his/her intent and interests. In order to achieve this goal, we have moved from considering the query-document matching to consider the user context. In fact, the user profile has been considered, in the literature, as the most important contextual element which can improve the accuracy of the search. It is integrated into the process of information retrieval in order to improve the user experience while searching for specific information. As time factor has gained increasing importance in recent years, the temporal dynamics are introduced to study the user profile evolution that consists mainly in capturing the changes of the user behavior, interests and preferences, and updating the profile accordingly. Prior work used to discern short-term and long-term profiles. The first profile type is limited to interests related to the user’s current activities while the second one represents user’s persisting interests extracted from his prior activities excluding the current ones. However, for users who are not very active, the short-term profile can eliminate relevant results which are more related to their personal interests. This is because their activities are few and separated over time. For users who are very active, the aggregation of recent activities without ignoring the old interests would be very interesting because this kind of profile is usually changing over time. Unlike those approaches, we propose, in this thesis, a generic time-sensitive user profile that is implicitly constructed as a vector of weighted terms in order to find a trade-off by unifying both current and recurrent interests. User profile information can be extracted from multiple sources. Among the most promising ones, we propose to use, on the one hand, searching history. Data from searching history can be extracted implicitly without any effort from the user and includes issued queries, their corresponding results, reformulated queries and click-through data that has relevance feedback potential. On the other hand, the popularity of Social Media makes it as an invaluable source of data used by users to express, share and mark as favorite the content that interests them. First, we modeled a user profile not only according to the content of his activities but also to their freshness under the assumption that terms used recently in the user’s activities contain new interests, preferences, and thoughts and should be considered more than old interests. In fact, many prior works have proved that the user interest is decreasing as time goes by. In order to evaluate the time-sensitive user profile, we used a set of data collected from Twitter, i.e a social networking and microblogging service. Then, we apply our re-ranking process to a Web search system in order to adapt the user’s online interests to the original retrieved results. Second, we studied the temporal dynamics within session search where recent submitted queries contain additional information explaining better the user intent and prove that the user hasn’t found the information sought from previous submitted ones. We integrated current and recurrent interactions within a unique session model giving more importance to terms appeared in recently submitted queries and clicked results. We conducted experiments using the 2013 TREC Session track and the ClueWeb12 collection that showed the effectiveness of our approach compared to state-of-the-art ones. Overall, in those different contributions and experiments, we prove that our time-sensitive user profile ensures better performance of personalization and helps to analyze user behavior in both session search and social media contexts.
Document type :
Complete list of metadatas

Cited literature [201 references]  Display  Hide  Download
Contributor : Ameni Kacem <>
Submitted on : Saturday, February 17, 2018 - 7:35:43 PM
Last modification on : Thursday, October 24, 2019 - 2:44:12 PM


Files produced by the author(s)




  • HAL Id : tel-01707423, version 2



Ameni Kacem. Personalized Information Retrieval based on Time-Sensitive User Profile. Information Retrieval [cs.IR]. Université Paul Sabatier (Toulouse 3), 2017. English. ⟨tel-01707423⟩



Record views


Files downloads