Browsing by Author "Chaa, Messaoud"

Now showing 1 - 6 of 6
  • Thumbnail Image
    Item
    Apprentissage d’ordonnancements en recherche d’information structurée
    (2013-03-23) Chaa, Messaoud
    L’adoption accrue de XML comme format standard pour représenter les documents structurés nécessite le développement des systèmes, efficients et efficaces, capable de retrouver les éléments XML pertinents à une requête d'utilisateur. Ces éléments sont ensuite présentés ordonnés en fonction de leur pertinence par rapport à la requête. Généralement la stratégie adoptée consiste à combiner plusieurs sources de pertinences dans une seule fonction de score et le poids de chaque source est donné manuellement selon des méthodes empiriques. Il est connu que, dans la recherche d’information classique, compte tenu de plusieurs sources de pertinences et l’utilisation des méthodes d’apprentissage d’ordonnancement en combinant ces sources de pertinences, améliore la performance des systèmes de recherche d’information. Dans ce travail, certains caractéristiques de pertinences des éléments XML, ont été définies et utilisées pour l’apprentissage d’ordonnancement dans les documents structurés. Notre objectif est de combiner ces caractéristiques afin d’obtenir la bonne fonction d’ordonnancement et montrer l’impact de chaque caractéristique dans la pertinence de l’élément XML. Des expérimentations sur une grande collection de la compagne d’évaluation de la recherche d’information XML (INEX) ont montré la performance de notre approche.
  • Thumbnail Image
    Item
    CERIST at INEX 2015: Social Book Search Track
    (CERIST, 2015) Chaa, Messaoud; Nouali, Omar
    In this paper, we describe our participation in the INEX 2015 Social Book Search Suggestion Track (SBS). We have exploited in our experiments only the tags assigned by users to books provided from LibraryThing (LT). We have investigated the impact of the weight of each term of the topic in the retrieval model using two methods. In the first method, we have used the TF-IQF formula to assign a weight to each term of the topic. In the second method, we have used Rocchio algorithm to expand the query and calculate the weight of the tags assigned to the example books mentioned in the book search request. Parameters of our models have been tuned using the topics of INEX 2014 and tested on INEX 2015 Social Book Search track.
  • Thumbnail Image
    Item
    Combining Tags and Reviews to Improve Social Book Search Performance
    (Springer, 2018-08-15) Chaa, Messaoud; Nouali, Omar; Bellot, Patrice
    The emergence of Web 2.0 and social networks have provided important amounts of information that led researchers from different fields to exploit it. Social information retrieval is one of the areas that aim to use this social information to improve the information retrieval performance. This information can be textual, like tags or reviews, or non textual like ratings, number of likes, number of shares, etc. In this paper, we focus on the integration of social textual information in the research model. As it seems logical that integrating tags in the retrieval model should not be in the same way taken to integrate reviews, we will analyze the different influences of using tags and reviews on both the settings of retrieval parameters and the retrieval effectiveness. After several experiments, on the CLEF social book search collection, we concluded that combining the results obtained from two separate indexes and two models with specific parameters for tags and reviews gives good results compared to when using a single index and a single model.
  • Thumbnail Image
    Item
    Learning to Rank in XML Information Retrieval: Which Feature Improve the Best?
    (IEEE, 2012-08-23) Chaa, Messaoud; Nouali, Omar; Bal, Kamal
    The augmented adoption of XML as the standard format for representing a document structure requires the development of tools to retrieve and rank effectively elements of the XML documents. It’s known that in information retrieval, considering multiple sources of relevance improves information retrieval. In this work some relevance features are defined and used in a learning to rank approach for XML information retrieval. Our aim is to combine theses features to derive good ranking function and show the impact of each feature in the relevance of XML element. Experiments on a large collection from the XML Information Retrieval evaluation campaign (INEX) showed good performance of the approach.
  • Thumbnail Image
    Item
    New Technique to Deal With Verbose Queries in Social Book Search
    (CERIST, 2017) Chaa, Messaoud; Nouali, Omar; Bellot, Patrice
    Verbose query reduction and query term weighting are automatic techniques to deal with verbose queries. The objective is either to assign an appropriate weight to query terms according to their importance in the topic, or outright remove unsuitable terms from the query and keep only the suitable terms to the topic and user’s need. These techniques improve performance and provide good results for ad hoc information retrieval. In this paper we propose a new approach to deal with long verbose queries in Social Information Re-trieval (SIR) by taking Social Book Search as an example. In this approach, a new statistical measure was introduced to reduce and weight terms of verbose queries. Next, we expand he query by exploiting the similar books mentioned by users in their queries. We find that the proposed approach improves significantly the results.
  • Thumbnail Image
    Item
    Verbose Query Reduction by Learning to Rank for Social Book Search Track
    (CERIST, 2016-07) Chaa, Messaoud; Nouali, Omar; Bellot, Patrice
    In this paper, we describe our participation in the INEX 2016 Social Book Search Suggestion Track (SBS). We have exploited machine learning techniques to rank query terms and assign an appropriate weight to each one before applying a probabilistic information retrieval model (BM15). Thereafter, only the top-k terms are used in the matching model. Several features are used to describe each term, such as statistical features, syntactic features and others features like whether the term is present in similar books and in the profile of the topic starter. The model was learned using the 2014 and 2015 topics and tested with the 2016 topics. Our experiments show that our approach improves the search results.