Rafael Muñoz Guillena

Analysing an Approach to Information Retrieval of Visual Descriptions with IR-n, a System Based on Passages

This paper analyses an approach made to the development of a textual image retrieval system by the University of Alicante using IR-n, a text-based information retrieval (IR) system. With only a minimal amount of adaptations to the features of this task, our system has obtained precision results above the mean average of participants for ImageCLEF07 both for English (0.1604 vs 0.1388) and for Spanish (0.1482 vs 0.1450). For German, our results were below the mean (0.0991 vs 0.1331), which could be due to the fact that our system does not incorporate a splitter for the treatment of this agglutinative language. These results are obtained without the incorporation of image recovery dominion techniques. The error analysis shows us that there is still a considerable amount of work to do concerning text-based techniques in order to improve the system, but also shows that the key to successful participation in this task is to mix text and image resources.

The Weighting Factors to Improve Predictability on Twitter

The result of the analysis of a thematic in a social network is to find a measure that allows the principal actors to know their performance, that is, they can define or maintain strategies and courses of action in order to optimize their communication. It is necessary to formally define the principles of analysis in Social Networks in order to use their characteristics better and to be able to contextualize the concept and use of weighting factors to improve their predictability. When Social Networks are going to be used as a mechanism to predict social behavior, for example, to predict the outcome of a political election, weighting factors must be introduced to try to match the data collected from the Social Network with those of a sample. In this article we have defined the methodology to incorporate the geographic weighting factors and several formulas have been created that allow reprocessing the data downloaded from Twitter in which its polarity has been determined by classical NLP methods to increase the predictive power.