UAlacant: Using Online Machine Translation for Cross-Lingual Textual Entailment

First Joint Conference on Lexical and Computational Semantics (*SEM), pages 472–476, Montr´eal, Canada, June 7-8, 2012. c⃝2012 Association for Computational Linguistics UAlacant: Using Online Machine Translation for Cross-Lingual Textual Entailment Miquel Espl`a-Gomis and Felipe S´anchez-Mart´ınez and Mikel L. Forcada Departament de Llenguatges i Sistemes Inform`atics Universitat d’Alacant, E-03071 Alacant, Spain {mespla,fsanchez,mlf}@dlsi.ua.es Abstract This paper describes a new method for cross- lingual textual entailment (CLTE) detection based on machine translation (MT). We use sub-segment translations from different MT systems available online as a source of cross- lingual knowledge. In this work we describe and evaluate different features derived from these sub-segment translations, which are used by a support vector machine classifier to detect CLTEs. We presented this system to the Se- mEval 2012 task 8 obtaining an accuracy up to 59.8% on the English–Spanish test set, the sec- ond best performing approach in the contest. 1 Introduction Cross-lingual textual entailment (CLTE) detec- tion (Mehdad et al., 2010) is an extension of the textual entailment (TE) detection (Dagan et al., 2006) problem. TE detection consists of finding out, for two text fragments T and H in the same language, whether T entails H from a semantic point of view or not. CLTE presents a similar problem, but with T and H written in different languages. During the last years, many authors have focused on resolving TE detection, as solutions to this prob- lem have proved to be useful in many natural lan- guage processing tasks, such as question answer- ing (Harabagiu and Hickl, 2006) or machine trans- lation (MT) (Mirkin et al., 2009; Pad´o et al., 2009). Therefore, CLTE may also be useful for related tasks in which more than one language is involved, such as cross-lingual question answering or cross-lingual information retrieval. Although CLTE detection is a relatively new problem, it has already been tack- led. Mehdad et al. (2010) propose to use machine translation (MT) to translate H from LH, the lan- guage of H, into LT , the language of T, and then use any of the state-of-the-art TE approaches. In a later work (Mehdad et al., 2011), the authors use MT, but in a more elaborate way. They train a phrase-based statistical MT (PBSMT) system (Koehn et al., 2003) translating from LH to LT , and use the translation table obtained as a by-product of the training process to extract a set of features which are processed by a support vector machine classifier (Theodoridis and Koutroumbas, 2009, Sect. 3.7) to decide whether T entails H or not. Castillo (2011) discusses another machine learning approach in which the features are obtained from semantic similarity measures based on WordNet (Miller, 1995). In this work we present a new approach to tackle the problem of CLTE detection using a machine learn- ing approach, partly inspired by that of Mehdad et al. (2011). Our method uses MT as a source of infor- mation to detect semantic relationships between T and H. To do so, we firstly split both T and H into …

Content truncated to 500 words