Rafael M. Terol

IR-n r2: using normalized passages

This paper describes the fourth participation of the IR-n system (Alicante University) at the CLEF evaluation campaigns. For CLEF 2004, we modified our similarity measure and query expansion model. For the similarity measure, we now use normalization based on the number of words for each of the passages. We tested two different approaches for query expansion: the first one is based on documents and the second on passages.

Deep vs. shallow semantic analysis applied to textual entailment recognition

This paper covers two different methods of recognising entailment between the text/hypothesis pair by processing logic forms. These two methods are based on knowledge sources. The logic forms of both the text and the hypothesis are inferred by analysing the syntactic dependency relationships between their words. Both approaches use the WordNet lexical database as knowledge source and obtain a semantic similarity score by means of WordNet relations. The difference between them is the treatment of these relations. Whereas one method carries out a deeper analysis considering many WordNet relations, the other one is shallower and manages only a reduced number of relations. These two approaches have been evaluated using the PASCAL Second RTE Challenge data and evaluation methodology.

An application of NLP rules to spoken document segmentation task

One of the main differences between Spoken Document Retrieval (SDR) systems and Text Retrieval systems is the need of a segmentation process that detects the story boundaries. However, until now, SDR researchers have not paid attention in building correct segments more than considering slidding windows of a fixed size in time. In this paper, new methodology for evaluating segments to SDR task, and the evaluation of three possible strategies are presented over the TREC-9 SDR collection. Moreover, the justification of each strategy is discussed.

An Architecture for Spoken Document Retrieval

In this paper main features of spoken document retrieval process are described. Cross language spoken document retrieval (CL-SDR) systems features which participated in the last CL-SDR track at CLEF conference are also showed. Following, we describe our CL-SDR system architecture and its features with the aim to produce better results than obtained by showed systems.

User behaviour and lexical ambiguity in cross-language image retrieval

Our objective in this paper is to determine the necessity of Word Sense Disambiguation in Information Retrieval tasks, according to user behaviour. We estimate and analyse the lexical ambiguity of queries in Cross-Language Image Retrieval (using search logs from a multilingual search interface for Flickr) and measure its correlation with search effectiveness. We show to what extent the lexical ambiguity of a query can influence the successful retrieval of an image in a multilingual framework.

Applying NLP techniques and biomedical resources to medical questions in QA performance

Nowadays, there is an increasing interest in research on QA over restricted domains. Concretely, in this paper we will show the process of question analysis in a medical QA system. This system is able to obtain answers to different natural language questions according to a question taxonomy. In this system we combine the use of NLP techniques and biomedical resources. The main NLP technique is the use of logic forms and the pattern matching technique in this question analysis performance.

The university of alicante at CL-SR track

In this paper, the new features that IR-n system applies on the topic processing for CL-SR are described. This set of features are based on applying logic forms to topics with the aim of incrementing the weight of topic terms according to a set of syntactic rules.

A knowledge-based textual entailment approach applied to the AVE task

We propose a system that has been initially used for Recognising Textual Entailment (RTE). Due to the fact that these two tasks (AVE and RTE) rely on the same main idea, our system was considered appropriate to be applied to AVE. Our system represents texts by means of logic forms and computes the semantic similarity between them. We have also designed a voting strategy between our system and the MLEnt system also developed at the University of Alicante. Although the results have not been very high, we consider them quite promising.

Applying Logic Forms and Statistical Methods to CL-SR Performance

This paper describes a CL-SR system that employs two different techniques: the first one is based on NLP rules that consist on applying logic forms to the topic processing while the second one basically consists on applying the IR-n statistical search engine to the spoken document collection. The application of logic forms to the topics allows to increase the weight of topic terms according to a set of syntactic rules. Thus, the weights of the topic terms are used by IR-n system in the information retrieval process.

Bilingual and multilingual experiments with the IR-n system

Our paper describes the participation of the IR-n system at CLEF-2005. This year, we participated in the bilingual task (English-French and English-Portuguese) and the multilingual task (English, French, Italian, German, Dutch, Finish and Swedish). We introduced the method of combined passages for the bilingual task. Futhermore we have applied the method of logic forms in the same task. For the multilingual task we had a joint participation with the University of Alicante and University of Jaen. We want to emphasize the good score achieved in the bilingual task improving around 45% in terms of average precision.