Inference for Natural Language

Research output: Chapter in Book/Report/Conference proceedingConference contribution

  • Authors:
  • Amal Alshahrani
  • Allan Ramsay

Abstract

The main aim of this study is to develop a naturallanguage inference (NLI) engine that ismore robust than typical systems that are basedon post-Montague approaches to semantics andmore accurate than the kinds of shallow approachesusually used for textual entailment, Theterm robustness is concerned with processing asmany inputs as possible successfully, and theterm accuracy is concerned with producing correctresult. In recent years, several approacheshave been proposed for NLI. These approachesrange from shallow approaches to deep approaches.However, each approach has a numberof limitations, which we discuss in this paper. Weargue that all approaches to NLI share a commonarchitecture, and that it may be possible to overcomethe limitations inherent in the existing approachesby combining elements of both kindsof strategy.

Bibliographical metadata

Original languageEnglish
Title of host publicationProceedings of the Joint Symposium on Semantic Processing. Textual Inference and Structures in Corpora
EditorsOctavian Popescu, Alberto Lavelli, Octavian Popescu
Place of PublicationTrentino – Italy
PublisherACL Anthology
Pages60-64
Number of pages5
Publication statusPublished - 20 Nov 2013
EventJSSP2013 - Joint Symposium on Semantic Processing. Textual Inference and Structures in Corpora - Trentino – Italy
Event duration: 20 Nov 201322 Nov 2013

Conference

ConferenceJSSP2013 - Joint Symposium on Semantic Processing. Textual Inference and Structures in Corpora
CityTrentino – Italy
Period20/11/1322/11/13