Possibilistic Model for Relevance Feedback in Collaborative Information Retrieval
Fatiha Naouar, Lobna Hlaoua, Mohamed Nazih Omri · Int. J. Web Appl. · 2012
Abstract — Web information is too heterogeneous that users have difficulties to retrieve their needed information: text, image or video. In this context, the collaborative work presents one solution proposed to solve this problem. Collaborative retrieval enables the retrieval histories’ sharing between users having the same profile across multiple tools such as annotations. We propose in this paper to improve collaborative retrieval performance, considering the annotations as a new source of information describing documents. In our contribution, we propose to apply the relevance feedback to extend the user’s query. So we use a possibilistic approach to extract the relevant terms from annotations given in semi-structured documents returned by collaborative retrieval systems. Keywords-component; Collaborative retrieval system; possibility theory; annotation; relevance feedback. I. I NTRODUCTION Facing the vast mass of information found on the web today, a collaborative work’s necessary to help the user find his needs. This evolution has improved the performance of retrieval especially for the number of relevant information found and the time put to perform the retrieval. In effect, Working in collaboration allows you to partake the search history as well as formulate a query for collaboration. The collaborative work can be done through multiple tools, in particular considering the annotations which represent relevant information in relation to the document that are to allocate a collection of keywords. In spite of this collaborative framework, the user usually suffers when searching the information to satisfy his needs, which is usually poorly expressed through his query which is composed of simple keywords due to his modest knowledge. In this framework, we suggest to improve the performances of the collaborative retrieval by applying the relevance feedback to enrich the original query. This technique, which consist in extracting terms, starting from documents considered relevant and consider them in a new extended query, was already applied in classic IR [21] and in semi-structured Information retrieval [22] [11] and showed its interest. In our contribution we consider annotations as a new source of information since it allows description of the document by personal users’ judgments. The annotation is relatively relevant since it can be made by specialists or not-specialist.So the relevance feedback using annotations in a collaborative frame brings us back to resolve principally two problems: to known the choice of annotations which can be judged as valid data to consider and the retrieval of the relevant terms which can be re-injected to extend the query.Several retrieval works were interested in the validation of annotations. We focused this work on the retrieval of the relevant terms used in the annotations to be valid. To do it, we propose a possibilistic model for express the necessity and the possibility of relevance of the terms to be extracted. We present in the following section a related work on the methods developed for a better collaborative retrieval. We describe our possibilistic model for the retrieval of information in section 3. Then in the section 4 we represent experimentation and results and we conclude at the end. II. R