Building a Greek corpus for Textual Entailment

Abstract

The paper reports on completed work aimed at the creation of a resource, namely, the Greek Textual Entailment Corpus (GTEC) that is appropriate for guiding training and evaluation of a system that recognizes Textual Entailment in Greek texts. The corpus of textual units was collected in view of a range of NLP applications, where semantic interpretation is of paramount importance, and it was manually annotated at the level of Textual Entailment. Moreover, a number of linguistic annotations were also integrated that were deemed useful for prospect system developers. The critical issue was the development of a final resource that is re-usable and adaptable to different NLP systems, in order to either enhance their accuracy or to evaluate their output. We are hereby focusing on the methodological issues underpinning data selection and annotation. An initial approach towards the development of a system catering for the automatic Recognition of Textual Entailment in Greek is also presented and preliminary results are reported.

Resources

Details