Catherine Berrut

A Comparison of Deep Learning Based Query Expansion with Pseudo-Relevance Feedback and Mutual Information

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 a large amount of unstructured text 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. Experiments conducted on four CLEF collections show a statistically significant improvement over the language models and other expansion models.

Wikipedia-based semantic query enrichment

We deal, in this paper, with the short queries (containing one or two words) problem. Short queries have no sufficient information to express their semantics in a non ambiguous way. Pseudo-relevance feedback (PRF) approach for query expansion is useful in many Information Retrieval (IR) tasks. However, this approach does not work well in the case of very short queries. Therefore, we present instead of PRF a semantic query enrichment method based on Wikipedia. This method expands short queries by semantically related terms extracted from Wikipedia. Our experiments on cultural heritage corpora show significant improvement in the retrieval performance.

The Effective Relevance Link between a Document and a Query

This paper proposes to understand the retrieval process of relevant documents against a query as a two-stage process: at first an identification of the reason why a document is relevant to a query that we called the Effective Relevance Link, and second the valuation of this link, known as the Relevance Status Value (RSV). We present a formal definition of this semantic link between d and q. In addition, we clarify how an existing IR model, like Vector Space model, could be used for realizing and integrating this formal notion to build new effective IR methods. Our proposal is validated against three corpuses and using three types of indexing terms. The experimental results showed that the effective link between d and q is very important and should be more taken into consideration when setting up an Information Retrieval (IR) Model or System. Finally, our work shows that taking into account this effective link in a more explicit and direct way into existing IR models does improve their retrieval performance.

An Image Retrieval System Based on the Visualization of System Relevance via Documents

This paper describes a system for an image retrieval system in which relevance related and system-use related user strategies can be performed. The query supports, in addition to classical topical inputs, strategic parameters that can be set by the user, either directly or via the visualization of retrieved images that are organized with respect to which relevance criteria are verified. Typical retrieval situations are defined, that account for the dynamic aspect of a given retrieval session.

Revisiting Exhaustivity and Specificity Using Propositional Logic and Lattice Theory

Exhaustivity and Specificity in logical Information Retrieval framework were introduced by Nie [16]. However, even with some attempts, they are still theoretical notions without a clear idea of how to be implemented. In this study, we present a new approach to deal with them. We use propositional logic and lattice theory in order to redefine the two implications and their uncertainty P(d → q) and P(q → d). We also show how to integrate the two notions into a concrete IR model for building a new effective model. Our proposal is validated against six corpora, and using two types of terms (words and concepts). The experimental results showed the validity of our viewpoint, which state: the explicit integration of Exhaustivity and Specificity into IR models will improve the retrieval performance of these models. Moreover, there should be a type of balance between the two notions.

Revisiting the Term Frequency in Concept-Based IR Models

Indexing documents and queries using concepts, instead of word-based indexing, is an alternative approach, and it supposes to give a more meaningful indexing. However, this way of indexing needs to revisit some hypotheses of classical Information Retrieval. Therefore, we propose a new concept weighting approach, namely Relative Weight, which weights concepts with respect to their corresponding text in the documents or queries. In other words, it assigns to each concept a relative weight with respect to the other concepts in the same context. We explore interesting experimental results of our new weighting approach, compared to the classical approaches, through studying the retrieval performance of some classical IR models.

The PRIME Information Retrieval System Applied on a Medical Corpus

PRIME is a precision oriented information retrieval system, managing multimedia structured documents. An information retrieval system provides access to documents based on their semantic content. The precision of such a system is measured as its capacity to give not necessary all, but only good answers. An information retrieval system is precision oriented when the information need of the users requires precise answers from the system; this leads the users to give also precise queries. Such systems are based on complex and controled representation languages, allowing a deep semantic understanding of the documents and the queries. The precision of an information retrieval system is quantified when instanciating it onto an application and a category of users. This paper describes the PRIME system, as we applied it on a medical corpus of documents for specialist physicians. The structured corpus contains medical texts, radiologic images and medical records. The software architecture of PRIME is built on the O2 system; the data structures dealing with the retrieval are conceptual graphs; the interfaces are developed in X-Motif and HTML using forms.

Hybrid query expansion model for text and microblog information retrieval

Query expansion (QE) is an important process in information retrieval applications that improves the user query and helps in retrieving relevant results. In this paper, we introduce a hybrid query expansion model (HQE) that investigates how external resources can be combined to association rules mining and used to enhance expansion terms generation and selection. The HQE model can be processed in different configurations, starting from methods based on association rules and combining it with external knowledge. The HQE model handles the two main phases of a QE process, namely: the candidate terms generation phase and the selection phase. We propose for the first phase, statistical, semantic and conceptual methods to generate new related terms for a given query. For the second phase, we introduce a similarity measure, ESAC, based on the Explicit Semantic Analysis that computes the relatedness between a query and the set of candidate terms. The performance of the proposed HQE model is evaluated within two experimental validations. The first one addresses the tweet search task proposed by TREC Microblog Track 2011 and an ad-hoc IR task related to the hard topics of the TREC Robust 2004. The second experimental validation concerns the tweet contextualization task organized by INEX 2014. Global results highlighted the effectiveness of our HQE model and of association rules mining for QE combined with external resources.

L'expressivité des modèles de recherche d'informations précises : Le support de vocabulaires et son application à la recherche d'information médicale

Nous proposons dans cet article de modeliser l’expressivite des systemes de recherche d’information (SRI). En effet peu de cadres de modelisation sont disponibles pour specifier les SRIs. Nous proposons un tel cadre sur lequel nous portons un interet particulier a la modelisation de l’expressivite, c’est-a-dire a ce que ce modele est capable de decrire. Le niveau d’expressivite est important dans les SRIs, et positionner un systeme au bon niveau permet d’obtenir de meilleurs resultats. Le cadre de modelisation que nous proposons permet ainsi de choisir l’expressivite d’un modele et de comparer des modeles sur leur niveau d’expressivite. Nous montrons en dernier lieu les possibilites offertes par ce cadre pour selectionner le niveau d’expressivite sur une tâche de recherche d’information medicale, ou les utilisateurs expriment des besoins complexes.