JU_CSE_NLP: Language Independent Cross-lingual Textual Entailment System

First Joint Conference on Lexical and Computational Semantics (*SEM), pages 689–695, Montr´eal, Canada, June 7-8, 2012. c⃝2012 Association for Computational Linguistics JU_CSE_NLP: Language Independent Cross-lingual Textual Entailment System Snehasis Neogi1, Partha Pakray2, Sivaji Bandyopadhyay1, Alexander Gelbukh3 1Computer Science & Engineering Department Jadavpur University, Kolkata, India 2Computer Science & Engineering Department Jadavpur University, Kolkata, India Intern at Xerox Research Centre Europe Grenoble, France 3Center for Computing Research National Polytechnic Institute Mexico City, Mexico {snehasis1981,parthapakray}@gmail.com sbandyopadhyay@cse.jdvu.ac.in gelbukh@gelbukh.com Abstract This article presents the experiments car- ried out at Jadavpur University as part of the participation in Cross-lingual Textual Entailment for Content Synchronization (CLTE) of task 8 @ Semantic Evaluation Exercises (SemEval-2012). The work ex- plores cross-lingual textual entailment as a relation between two texts in different lan- guages and proposes different measures for entailment decision in a four way clas- sification tasks (forward, backward, bidi- rectional and no-entailment). We set up different heuristics and measures for eva- luating the entailment between two texts based on lexical relations. Experiments have been carried out with both the text and hypothesis converted to the same lan- guage using the Microsoft Bing translation system. The entailment system considers Named Entity, Noun Chunks, Part of speech, N-Gram and some text similarity measures of the text pair to decide the en- tailment judgments. Rules have been de- veloped to encounter the multi way entailment issue. Our system decides on the entailment judgment after comparing the entailment scores for the text pairs. Four different rules have been developed for the four different classes of entailment. The best run is submitted for Italian – English language with accuracy 0.326. 1 Introduction Textual Entailment (TE) (Dagan and Glick- man, 2004) is one of the recent challenges of Natural Language Processing (NLP). The Task 8 of SemEval-20121 [1] defines a textual en- tailment system that specifies two major as- pects: the task is based on cross-lingual corpora and the entailment decision must be four ways. Given a pair of topically related text fragments (T1 and T2) in different languages, the CLTE task consists of automatically anno- tating it with one of the following entailment judgments: i. Bidirectional (T1 ->T2 & T1 <- T2): the two fragments entail each other (semantic equiva- lence) ii. Forward (T1 -> T2 & T1!<- T2): unidirec- tional entailment from T1 to T2 . iii. Backward (T1! -> T2 & T1 <- T2): unidirec- tional entailment from T2 to T1. iv. No Entailment (T1! -> T2 & T1! <- T2): there is no entailment between T1 and T2. CLTE (Cross Lingual Textual Entailment) task consists of 1,000 CLTE dataset pairs (500 for 1http://www.cs.york.ac.uk/semeval2012/index.php?id=tasks 689 training and 500 for test) available for the fol- lowing language combinations: - Spanish/English (spa-eng) - German/English (deu-eng). - Italian/English (ita-eng) - French/English (fra-eng) Seven Recognizing Textual Entailment (RTE) evaluation tracks have already been held: RTE-1 in 2005 [2], RTE-2 [3] in 2006, RTE-3 [4] in 2007, RTE-4 [5] in 2008, RTE-5 [6] in 2009, RTE-6 [7] in 2010 and RTE-7 [8] in 2011. RTE …

Content truncated to 500 words