Extraction of Semantic Relation Based on Feature Vector from Wikipedia

In this paper, we propose a feature vector to extract semantic relations using dependency tree and parse tree. We exploit relation descriptions from infoboxes on Wikipedia documents. The features include part-of-speech, phrase label in dependency tree, and grammatical structure of phrase label, path of phrase label inherent in parse tree. In our experi ments, support vector machine and k-nearest neighbor are applied to extract relations from Wikipedia documents.