Catálogo de publicaciones - libros

Users of information retrieval (IR) systems require an interface that is powerful and easy-to-use in order to fulfill their information requirement. In XML-IR systems this is a non-trivial task since users expect these systems to fulfill both their structural and content requirements. Most existing XML-IR systems accept queries formatted in formal query languages, however, these languages are difficult to use. This paper presents NLPX – an XML-IR system with a natural language interface that is user friendly enough so it can be used intuitively, but sophisticated enough to be able to handle complex structured queries. NLPX accepts English queries that contain both users’ content and structural requirements. It uses a set of grammar templates to derive the structural and content requirements and translates them into a formal language (NEXI). The formal language queries can then be processed by many existing XML-IR systems. The system was developed for participation in the NLP Track of the INEX 2004 Workshop, and results indicated that natural language interfaces are able to capture users’ structural and content requirements, but not as accurately as some formal language interfaces.

This article presents the contribution of the “École Nationale Supérieure des Mines de Saint-Etienne (France)” to the new Natural Language Processing special Track of the third Initiative for Evaluation of XML Retrieval (INEX 2004). It discusses the place of NLP in XML retrieval and presents a method to analyse natural language queries.

An interactive track was included in INEX for the first time in 2004. The main aim of the track was to study the behaviour of searchers when interacting with components of XML documents. In this paper, we describe the motivation and aims of the track in detail, we outline the methodology and we present some initial findings from the analysis of the results.

The aim of this study was to provide fundamental data on users’ searching behavior when they are interacting with structured XML documents, and to discover the correlation between users’ search characteristics and their attitude towards evaluating the interactive XML IR system. 8 subjects took part in this experiment and we followed the experimental guidelines from the INEX 2004 Interactive Track organizers with no modification. In total, 16 transaction logs from the search sessions and 6 responses from the questionnaires per subject were collected and analyzed to achieve our research goals. Collected data was analyzed using SPSS Windows 10.0 and the results are presented in various tables. No significant correlation was found between the users’ chosen search characteristics and their attitude towards evaluating the XML IR system in this experiment.