Support vector machines for semantic relation extraction in Spanish language

Abstract

Relation Extraction (RE) is one of the most important topics in NLP (Natural Language Processing). Many tasks such as semantic relation extraction, sentiment analysis, opinion mining, question answering systems and text summarization are supported by RE. The aim of this paper is to present a semantic relations classifier in which are incorporate lexical features, named entity features and syntactic structures. Relations between two entities are classified based on the Datasets for Generic Relation Extraction (reACE). We translate the reACE corpus to the Spanish language for all relation types and subtypes. The results shows a F-score of 75.25%, it is a significant improvement of 11.5% over the baseline model. Finally, we discuss the results according to the model and the useful information to support the forecasting process.