Search

Recently, different works on bilingual lexicon extraction from comparable corpora have been proposed. This paper presents how to combine differents methods for bilingual lexicon extraction based on standard context vectors and advanced text mining methods. In this respect, we focus on combining bilingual lexicons based on context vectors, association rules and contextual meta-rules. The combination of lexicons leads to a less sparse representation in order to extract the most effective translations from these lexicons and create an optimal bilingual lexicon. An experimental validation conducted on two pairs of languages of the CLEF 2003 campaign evaluation, shows that the combination of the models give a significant improvement compared to the standard approach.

International audience; Logic-based information retrieval (IR) models represent the retrieval decision as an implication d→q between a document d and a query q, where d and q are logical sentences. However, d→q is generally a binary decision, thus we need a measurement to estimate the degree to which d implies q, denoted U(d→q). Most of the existing logic-based IR models either do not precisely define the implication d→q or use non-classical definitions. Some models also define the uncertainty U in informal ways. More importantly, they use two non-related frameworks to define d→q and its uncertainty U, even though the two notions are intrinsically related. The goal of this study is to propose a new logic-based IR model, which overcome these shortcomings. To this end, we first propose to replace the implication d→q by the validity of material implication ⊨d⊃q. Second, we redefine and adapt the mathematical relationship between logics, lattices and probability. Our new IR model presents a possible formalism for van Rijsbergen's intuition about replacing U(d→q) by P(q∣d).

International audience; Automatic query expansion techniques are widely applied for improving text retrieval performance, using a variety of approaches that exploit several data sources for finding expansion terms. Selecting expansion terms is challenging and requires a framework capable of extracting term relationships. Recently, several Natural Language Processing methods , based on Deep Learning, are proposed for learning high quality vector representations of terms from large amounts of unstructured text data with billions of words. These high quality vector representations capture a large number of term relationships. In this paper, we experimentally compare several expansion methods with expansion using these term vector representations. We use language models for information retrieval to evaluate expansion methods. The experiments are conducted on four CLEF collections show a statistically significant improvement over the language models and other expansion models.

International audience; Microblogging sites, like Twitter, have emerged as a popular platform for expressing opinions. Bound to 140 characters, Twitter's publications (tweets) are very short and not always written maintaining formal grammar and proper spelling. The spelling variations increase the likelihood of vocabulary mismatch and make the tweets dicult to understand without some kind of context. The task of tweet contextualization was organized around these issues. It aims to provide an automatic readable context explaining a given tweet, in order to help the reader understand this latter. In this paper, we describe statistical and semantic approaches for the tweet contextualization task. While the statistical one is based on association rules mining, the semantic one uses Wikipedia as an external knowledge source. The e↵ectiveness of our approaches is proved through an experimental study conducted on the INEX 2013 collection.

Most information retrieval systems rely on the strict equality of terms between document and query in order to retrieve relevant documents to a given query. The term mismatch problem appears when users and documents’ authors use different terms to express the same meaning. Statistical translation models are proposed as an effective way to adapt language models in order to mitigate term mismatch problem by exploiting semantic relations between terms. However, translation probability estimation is shown as a crucial and a hard practice within statistical translation models. Therefore, we present an alternative approach to statistical translation models that formally incorporates semantic relations between indexing terms into language models. Experiments on different CLEF corpora from the medical domain show a statistically significant improvement over the ordinary language models, and mostly better than translation models in retrieval performance. The improvement is related to the rate of general terms and their distribution inside the queries.