Barrier Features for Classification of Semantic Relations

Proceedings of Recent Advances in Natural Language Processing, pages 509–514, Hissar, Bulgaria, 12-14 September 2011. Barrier Features for Classification of Semantic Relations Anita Alicante University ”Federico II”, Naples, Italy anita.alicante@unina.it Anna Corazza University ”Federico II”, Naples, Italy corazza@na.infn.it Abstract Approaches based on machine learning, such as Support Vector Machines, are of- ten used to classify semantic relations be- tween entities. In such framework, classi- fication accuracy strongly depends on the set of features which are used to represent the input to the classifier. We are propos- ing here a new type of features, namely the barrier features, which can be used in ad- dition to more usual features, such as n- grams of PoS, word suffixes and prefixes, hypernyms from WordNet etc., and to the parse tree of the whole sentence. Barrier features aim at giving a compact repre- sentation of the context of each entity in- volved in the relation. The effectiveness of the new features is assessed on documents from the TREC data set annotated by Roth and Yih. The obtained results show not only that the performance of the proposed approach are state-of-the-art but also that such improvement is due to the introduc- tion of the barrier features. 1 Introduction Different approaches have been proposed for the identification of entities and relations in text. In this paper we focus only on the task of classifica- tion of semantic relations, that is of assigning to each semantic relation a label taken from a finite set, and therefore we assume that pairs of related entities are given us together with their labels. In general, not all relation labels are compatible with every pair of entity types. Table 3 reports all such constraints in the considered data set. Thus not all relation labels can be compatible with all en- tity pairs and therefore we decompose the prob- lem in several binary classifications, one for each possible relation label and apply Support Vector Machines (SVMs) to each of these subtasks. Moreover, by applying SVMs with a combi- nation of different kernel functions we can han- dle together different kinds of information, both structured and not. In fact, we applied tree ker- nels (Moschitti, 2006) to the whole sentence parse tree and a linear kernel to a vector of binary fea- tures extracted from the words surrounding each of the involved entities. Among the latter, we introduce a novel kind of binary feature, which we call barrier features and experimentally prove that they improve classification performance. Al- though inspired by the barrier rules of the con- straint grammar framework proposed in (Karlsson et al., 1995) for PoS tagging, barrier features have been completely redesigned for this task as binary values rather than rules. The experimental assessment of the approach is performed on the newswire documents annotated by Roth and Yih (Roth and Yih, 2004). The best published results for relation classification on this data set have been obtained by the MO|K system, described in (Giuliano et al., 2007). The follow- ing Section …

Content truncated to 500 words