Evaluación del rendimiento de los sistemas de búsqueda de respuestas de dominio general

Abstract

Information overload is felt more strongly on the Web than elsewhere. Question-answering systems (QA systems) are considered as an alternative to traditional information retrieval systems, because they give correct and understandable answers rather than just offering a list ofdocuments. Four answer search systems available online have been analyzed: START, QuALiM, SEMOTE, and TrueKnowledge. They were analyzed through a wide range of questions that prompted responses of definitions, facts, and closed lists pertaining to different thematic areas. The answers were analyzed using several specific measurements (MRR, TRR, FHS, MAP and precision). The results are encouraging and they show that these systems, although each one different, are potentially valid for precise information retrieval of diverse types and thematic areas. © 2013 CSIC.