L. A. Ureña-López

Query expansion with a medical ontology to improve a multimodal information retrieval system

Searching biomedical information in a large collection of medical data is a complex task. The use of tools and biomedical resources could ease the retrieval of the information desired. In this paper, we use the medical ontology MeSH to improve a Multimodal Information Retrieval System by expanding the users query with medical terms. In order to accomplish our experiments, we have used the dataset provided by ImageCLEFmed task organizers for years 2005 and 2006. This dataset is composed of a multimodal collection (images and text) of clinical cases, a list of queries for each year, and a list of relevance judgments for each query to evaluate the results. The results from the experiments show that the use of a medical ontology to expand the queries greatly improves the results.

Integrating MeSH Ontology to Improve Medical Information Retrieval

This paper describes the SINAI team participation in the ImageCLEFmed campaign. The SINAI research group has participated in the multilingual image retrieval subtask. The experiments accomplished are based on the integration of specific knowledge in the topics. We have used the MeSH ontology to expand the queries. The expansion consists in searching terms from the topic query in the MeSH ontology in order to add similar terms. We have processed the set of collections using Information Gain (IG) in the same way as in ImageCLEFmed 2006. In our experiments mixing visual and textual information we obtain better results than using only textual information. The weigth of the textual information is very strong in this mixed strategy. In the experiments with a low textual weight, the use of IG improves the results obtained.

The learning vector quantization algorithm applied to automatic text classification tasks

Automatic text classification is an important task for many natural language processing applications. This paper presents a neural approach to develop a text classifier based on the Learning Vector Quantization (LVQ) algorithm. The LVQ model is a classification method that uses a competitive supervised learning algorithm. The proposed method has been applied to two specific tasks: text categorization and word sense disambiguation. Experiments were carried out using the Reuters-21578 text collection (for text categorization) and the Senseval-3 corpus (for word sense disambiguation). The results obtained are very promising and show that our neural approach based on the LVQ algorithm is an alternative to other classification systems.

Using information gain to improve multi-modal information retrieval systems

Nowadays, access to information requires managing multimedia databases effectively, and so, multi-modal retrieval techniques (particularly images retrieval) have become an active research direction. In the past few years, a lot of content-based image retrieval (CBIR) systems have been developed. However, despite the progress achieved in the CBIR, the retrieval accuracy of current systems is still limited and often worse than only textual information retrieval systems. In this paper, we propose to combine content-based and text-based approaches to multi-modal retrieval in order to achieve better results and overcome the lacks of these techniques when they are taken separately. For this purpose, we use a medical collection that includes both images and non-structured text. We retrieve images from a CBIR system and textual information through a traditional information retrieval system. Then, we combine the results obtained from both systems in order to improve the final performance. Furthermore, we use the information gain (IG) measure to reduce and improve the textual information included in multi-modal information retrieval systems. We have carried out several experiments that combine this reduction technique with a visual and textual information merger. The results obtained are highly promising and show the profit obtained when textual information is managed to improve conventional multi-modal systems.

Crowd explicit sentiment analysis

With the rapid growth of data generated by social web applications new paradigms in the generation of knowledge are opening. This paper introduces Crowd Explicit Sentiment Analysis (CESA) as an approach for sentiment analysis in social media environments. Similar to Explicit Semantic Analysis, microblog posts are indexed by a predefined collection of documents. In CESA, these documents are built up from common emotional expressions in social streams. In this way, texts are projected to feelings or emotions. This process is performed within a Latent Semantic Analysis. A few simple regular expressions (e.g. “I feel X”, considering X a term representing an emotion or feeling) are used to scratch the enormous flow of micro-blog posts to generate a textual representation of an emotional state with clear polarity value (e.g. angry, happy, sad, confident, etc.). In this way, new posts can be indexed by these feelings according to the distance to their textual representation. The approach is suitable in many scenarios dealing with social media publications and can be implemented in other languages with little effort. In particular, we have evaluated the system on Polarity Classification with both English and Spanish data sets. The results show that CESA is a valid solution for sentiment analysis and that similar approaches for model building from the continuous flow of posts could be exploited in other scenarios.

Comparing Several Textual Information Retrieval Systems for the Geographical Information Retrieval Task

This paper presents a comparison between three different Information Retrieval (IR) systems employed in a particular Geographical Information Retrieval (GIR) system, the GeoUJA IR, a GIR architecture developed by the SINAI research group. It could be interesting and useful for determining which of the most used IR systems works better in GIR task. In the experiments, we have used the Lemur, Terrier and Lucene search engines using mono and bilingual queries. We present baseline cases, without applying any external processes, such as query expansion or filtering. In addition, we have used the default settings of each IR system. Results show that Lemur works better using monolingual queries and Terrier works better using the bilingual ones.

Selection strategies for multi-label text categorization

In multi-label text categorization, determining the final set of classes that will label a given document is not trivial. It implies first to determine whether a class is suitable of being attached to the text and, secondly, the number of them that we have to consider. Different strategies for determining the size of the final set of assigned labels are studied here. We analyze several classification algorithms along with two main strategies for selection: by a fixed number of top ranked labels, or using per-class thresholds. Our experiments show the effects of each approach and the issues to consider when using them.In multi-label text categorization, determining the final set of classes that will label a given document is not trivial. It implies first to determine whether a class is suitable of being attached to the text and, secondly, the number of them that we have to consider. Different strategies for determining the size of the final set of assigned labels are studied here. We analyze several classification algorithms along with two main strategies for selection: by a fixed number of top ranked labels, or using per-class thresholds. Our experiments show the effects of each approach and the issues to consider when using them.

Filtering for Improving the Geographic Information Search

This paper describes the GEOUJA System, a Geographical Information Retrieval (GIR) system submitted by the SINAI group of the University of Jaen in GeoCLEF 2007. The objective of our system is to filter the documents retrieved from an information retrieval (IR) subsystem, given a multilingual statement describing a spatial user need. The results of the experiments show that the new heuristics and rules applied in the geo-relation validator module improve the general precision of our system. The increasing of the number of documents retrieved by the information retrieval subsystem also improves the final results.

GEOUJA system the first participation of the University of Jaén at GeoCLEF 2006

This paper describes the first participation of the SINAI group of the University of Jaen in GeoCLEF 2006. We have developed a system made up of three main modules: the Translation Subsystem, that works with queries into Spanish and German against English collection; the Query Expansion subsystem, that integrates a Named Entity Recognizer, a thesaurus expansion module and a geographical information-gazetteer module; and the Information Retrieval subsystem. We have participated in the monolingual and the bilingual tasks. The results obtained shown that the use of geographical and thesaurus information for query expansion does not improve the retrieval in our experiments.

The university of jaén at ImageCLEF 2005: adhoc and medical tasks

In this paper, we describe our first participation in the ImageCLEF campaign. The SINAI research group participated in both the ad hoc task and the medical task. For the first task, we have used several translation schemes as well as experiments with and without Pseudo Relevance Feedback (PRF). A voting-based system has been developed, for the ad hoc task, joining three different systems of participant Universities. For the medical task, we have also submitted runs with and without PRF, and experiments using only textual query and using textual mixing with visual query.