Evaluación de sistemas españoles de recuperación de información distribuída en Internet

The set of information spaces collectively referred as Internet poses serious problems to information retrieval tasks. Content evolution of Internet spaces and documents is reviewed and distinctive features of web documents are empathized. Web search engines are classified according to their scope, functionalities and retrieval philosophy. A chapter is devoted to the characterization of Spanish web though the study of a random set of web sites, their quantitative composition and their qualitative features. The analysis of search engines of the Spanish web begins with a study of coverage, methods of crawling, data schema and indexing mechanisms. Finally, eight search engines (AltaVista, EnlaWeb, Lycos, Olé/Terra, Ozú, Sol, Ya and Yahoo) were evaluated in retrieving information from Spanish web space. Indicators chosen were their relative coverage, specific offering, proportion of dead links and accessibility of Spanish websites. Performance was determined by relative recall and precision in retrieval during the first quarter in 2003. Search topics and relevance of results were determined by the end users. 12,4% of the searches led to dead links and 76% of the pages were returned by only a single system. System performance, expressed in terms of recall ranged from 7% (AltaVista) to 14% (Ozú) and precision between 9% (Sol) and 30% (Ozú). Only Yahoo displayed typical inverse relationship between recall and precision figures. The rest of the systems invariably showed an increase in precision figures starting with the second or third search result, suggesting problems with the sorting algorithm