Towards Cross-Lingual Textual Entailment

Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the ACL, pages 321–324, Los Angeles, California, June 2010. c⃝2010 Association for Computational Linguistics Towards Cross-Lingual Textual Entailment Yashar Mehdad1,2, Matteo Negri1, Marcello Federico1 FBK-Irst1, University of Trento2 Trento, Italy {mehdad,negri,federico}@fbk.eu Abstract This paper investigates cross-lingual textual entailment as a semantic relation between two text portions in different languages, and pro- poses a prospective research direction. We argue that cross-lingual textual entailment (CLTE) can be a core technology for sev- eral cross-lingual NLP applications and tasks. Through preliminary experiments, we aim at proving the feasibility of the task, and provid- ing a reliable baseline. We also introduce new applications for CLTE that will be explored in future work. 1 Introduction Textual Entailment (TE) (Dagan and Glickman, 2004) has been proposed as a generic framework for modeling language variability. Given two texts T and H, the task consists in deciding if the meaning of H can be inferred from the meaning of T. So far, TE has been only applied in a monolingual setting, where both texts are assumed to be written in the same language. In this work, we propose and inves- tigate a cross-lingual extension of TE, where we as- sume that T and H are written in different languages. The great potential of integrating (monolingual) TE recognition components into NLP architectures has been reported in several works, such as ques- tion answering (Harabagiu and Hickl, 2006), infor- mation retrieval (Clinchant et al., 2006), informa- tion extraction (Romano et al., 2006), and document summarization (Lloret et al., 2008). To the best of our knowledge, mainly due to the absence of cross-lingual TE (CLTE) recognition components, similar improvements have not been achieved yet in any cross-lingual application. As a matter of fact, despite the great deal of attention that TE has received in recent years (also witnessed by five editions of the Recognizing Textual Entail- ment Challenge1), interest for cross-lingual exten- sions has not been in the mainstream of TE research, which until now has been mainly focused on the En- glish language. Nevertheless, the strong interest towards cross- lingual NLP applications (both from the market and research perspectives, as demonstrated by success- ful evaluation campaigns such as CLEF2) is, to our view, a good reason to start investigating CLTE, as well. Along such direction, research can now ben- efit from recent advances in other fields, especially machine translation (MT), and the availability of: i) large amounts of parallel and comparable corpora in many languages, ii) open source software to com- pute word-alignments from parallel corpora, and iii) open source software to set-up strong MT baseline systems. We strongly believe that all these resources can potentially help in developing inference mecha- nisms on multilingual data. Building on these considerations, this paper aims to put the basis for future research on the cross- lingual Textual Entailment task, in order to allow for semantic inference across languages in different NLP tasks. Among these, as a long-term goal, we plan …

Content truncated to 500 words