Abstract
The ambiguity problem is revealed when a query leads to more than one meaning. The query disambiguation process consists in enhancing the automatic understanding to provide diverse suggestions for the user according to his intent. Indeed, the query disambiguation generally requires knowledge structure which relates more than one entity to a given term. Moreover, since the ambiguity characterizes the human language, the integration of the user profile should be useful for the enhancement of this process. Indeed, it is usually carried out in the retrieval phase but recently it has been sparsely developed in the disambiguation of the query. In this paper, we propose a user-centric query disambiguation approach which implies two main steps: the Suggestions Mining and the Suggestions Personalization. The diversity of suggestions is extracted from the XML Wikipedia pages. Then, the mined suggestions are re-ranked according to the specifications of each user profile which allows the forecast of the most likely query expansion. A semantic graph is generated with the Suggestions-Preferences Mapping in order to illustrate the weighted relations between the provided suggestions and the preferences representing the user profile. Then, based on the preferences ranking and the best similarity scores obtained between the preferences and the suggestions, the Suggestions Re-Ranking process offers a personalized sort of the mined suggestions to achieve a personalized query disambiguation. We conduct an experimental study based on thirty ambiguous queries and some samples of user profiles. Promising results are achieved by the proposed approach.