BUAP: A First Approximation to Relational Similarity Measuring

First Joint Conference on Lexical and Computational Semantics (*SEM), pages 502–505, Montr´eal, Canada, June 7-8, 2012. c⃝2012 Association for Computational Linguistics BUAP: A First Approximation to Relational Similarity Measuring Mireya Tovar, J. Alejandro Reyes, Azucena Montes CENIDET, Department of Computer Science Int. Internado Palmira S/N, Col. Palmira Cuernavaca, Morelos, M´exico {mtovar, alexreyes06c, amr} @cenidet.edu.mx Darnes Vilari˜no, David Pinto, Saul Le´on B. Universidad Aut´onoma de Puebla, Faculty of Computer Science 14 Sur y Av. San Claudio, CU Puebla, Puebla, M´exico {darnes, dpinto}@cs.buap.mx saul.ls@live.com Abstract We describe a system proposed for measuring the degree of relational similarity beetwen a pair of words at the Task #2 of Semeval 2012. The approach presented is based on a vec- torial representation using the following fea- tures: i) the context surrounding the words with a windows size = 3, ii) knowledge ex- tracted from WordNet to discover several se- mantic relationships, such as meronymy, hy- ponymy, hypernymy, and part-whole between pair of words, iii) the description of the pairs with their POS tag, morphological informa- tion (gender, person), and iv) the average num- ber of words separating the two words in text. 1 Introduction The Task # 2 of Semeval 2012 focuses on measuring the degree of relational similarity between the ref- erence words pairs (training) and the test pairs for a given class (Jurgens et al., 2012). The training data set consists of 10 classes and the testing data set consists of the 69 classes. These datasets as well as the particularities of the task are better described at overview paper (Jurgens et al., 2012). In this paper we report the approach submit- ted to the competition, which is based on a vector space model representation for each pair (Salton et al., 1975). With respect to the type of features used, we have observed that Fabio Celli (Celli, 2010) con- siders that contextual information is useful, as well the lexical and semantic information are in the ex- traction of semantic relationships task. Additionally, in (Chen et al., 2010) and (Negri and Kouylekov, 2010) are proposed WordNet based features with the same purpose. In the experiments carried out in this paper, we use a set of lexical, semantic, WordNet-based and contextual features which allows to construct the vectors. Actually, we have tested a subset of the 20 contextual features proposed by Celli (Celli, 2010) and some of those proposed by Chen (Chen et al., 2010) and Negri (Negri and Kouylekov, 2010). The cosine similarity measure is used for deter- mining the degree of relational similarity (Frakes and Baeza-Yates, 1992) among the vectors. The rest of this paper is structured as follows. Section 2 describes the system employed. Section 3 show the obtained results. Finally, in Section 4 the final conclusions are given. 2 System description The approach reported in this paper measures the relational similarity of a set of word pairs that be- long to the same semantic relationship. Those word pairs are represented by means of the vector space model (Salton et al., 1975). …

Content truncated to 500 words