Cross-Lingual Textual Entailment and Applications

Mehdad, Yashar (2012) Cross-Lingual Textual Entailment and Applications. PhD thesis, University of Trento, Foundazione Bruno Kessler.

PDF (Cross-Lingual Textual Entailment and Applications) - Doctoral Thesis


Textual Entailment (TE) has been proposed as a generic framework for modeling language variability. The great potential of integrating (monolingual) TE recognition components into NLP architectures has been reported in several areas, such as question answering, information retrieval, information extraction and document summarization. Mainly due to the absence of cross-lingual TE (CLTE) recognition components, similar improvements have not yet been achieved in any corresponding cross-lingual application. In this thesis, we propose and investigate Cross-Lingual Textual Entailment (CLTE) as a semantic relation between two text portions in dierent languages. We present dierent practical solutions to approach this problem by i) bringing CLTE back to the monolingual scenario, translating the two texts into the same language; and ii) integrating machine translation and TE algorithms and techniques. We argue that CLTE can be a core tech- nology for several cross-lingual NLP applications and tasks. Experiments on dierent datasets and two interesting cross-lingual NLP applications, namely content synchronization and machine translation evaluation, conrm the eectiveness of our approaches leading to successful results. As a complement to the research in the algorithmic side, we successfully explored the creation of cross-lingual textual entailment corpora by means of crowdsourcing, as a cheap and replicable data collection methodology that minimizes the manual work done by expert annotators.

Item Type:Doctoral Thesis (PhD)
Doctoral School:Information and Communication Technology
PhD Cycle:XXIV
Subjects:Area 01 - Scienze matematiche e informatiche > INF/01 INFORMATICA
Repository Staff approval on:05 Jul 2012 11:52

Repository Staff Only: item control page