UMCC_DLSI: Multidimensional Lexical-Semantic Textual Similarity

First Joint Conference on Lexical and Computational Semantics (*SEM), pages 608–616, Montr´eal, Canada, June 7-8, 2012. c⃝2012 Association for Computational Linguistics UMCC_DLSI: Multidimensional Lexical-Semantic Textual Similarity Antonio Fernández, Yoan Gutiérrez, Alexander Chávez, Héctor Dávila, Andy González, Rainel Estrada , Yenier Castañeda DI, University of Matanzas Autopista a Varadero km 3 ½ Matanzas, Cuba Sonia Vázquez Andrés Montoyo, Rafael Muñoz, DLSI, University of Alicante Carretera de San Vicente S/N Alicante, Spain Abstract This paper describes the specifications and results of UMCC_DLSI system, which participated in the first Semantic Textual Similarity task (STS) of SemEval-2012. Our supervised system uses different kinds of semantic and lexical features to train classifiers and it uses a voting process to select the correct option. Related to the different features we can highlight the resource ISR-WN1 used to extract semantic relations among words and the use of different algorithms to establish semantic and lexical similarities. In order to establish which features are the most appropriate to improve STS results we participated with three runs using different set of features. Our best approach reached the position 18 of 89 runs, obtaining a general correlation coefficient up to 0.72. 1. Introduction SemEval 2012 competition for evaluating Natural Language Processing (NLP) systems presents a new task called Semantic Textual Similarity (STS) (Agirre et al., 2012). In STS the participating systems must examine the degree of semantic equivalence between two sentences. The goal of this task is to create a unified framework for the evaluation of semantic textual similarity modules and to characterize their impact on NLP applications. STS is related to Textual Entailment (TE) and Paraphrase tasks. The main difference is that STS 1 Integration of Semantic Resource based on WordNet. assumes bidirectional graded equivalence between the pair of textual snippets. In the case of TE the equivalence is directional (e.g. a student is a person, but a person is not necessarily a student). In addition, STS differs from TE and Paraphrase in that, rather than being a binary yes/no decision, STS is a similarity-graded notion (e.g. a student and a person are more similar than a dog and a person). This bidirectional gradation is useful for NLP tasks such as Machine Translation, Information Extraction, Question Answering, and Summarization. Several semantic tasks could be added as modules in the STS framework, “such as Word Sense Disambiguation and Induction, Lexical Substitution, Semantic Role Labeling, Multiword Expression detection and handling, Anaphora and Co-reference resolution, Time and Date resolution and Named Entity Recognition, among others”2 1.1. Description of 2012 pilot task In STS, all systems were provided with a set of sentence pairs obtained from a segmented corpus. For each sentence pair, s1 and s2, all participants had to quantify how similar s1 and s2 were, providing a similarity score. The output of different systems was compared to the manual scores provided by SemEval-2012 gold standard file, which range from 5 to 0 according to the next criterions3: • (5) “The two sentences are equivalent, as they mean the same thing”. 2 http://www.cs.york.ac.uk/semeval-2012/task6/ …

Content truncated to 500 words