Deep domain models for discourse analysis.

Leo Joskowicz, Tomasz Ksiezyk, Ralph Grishman

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


The authors address the problem of discourse analysis, and in particular finding causal relations between facts mentioned in messages, using a detailed domain model. This work is part of the PROTEUS (Prototype Text Understanding System) project, whose objective is to understand short narrative message about equipment installed in Navy ships. Casualty reports (CASREPs) describe failures of this equipment, together with maintenance actions performed by the crew on board. To capture the domain knowledge, the authors built a model of the equipment installed on the ship (initially, the starting air system) and demonstrated its use in several aspects of language understanding. The PROTEUS system has been substantially implemented and debugged and has been publically demonstrated operating on a small set of actual CASREPs.

Original languageEnglish (US)
Title of host publicationProc Annu AI Syst Gov Conf
EditorsJames H. Antonisse, John W. Benoit, Barry G. Silverman
Number of pages6
StatePublished - 1989
EventProceedings of the Annual AI Systems in Government Conference - Washington, DC, USA
Duration: Mar 27 1989Mar 31 1989

Publication series

NameProc Annu AI Syst Gov Conf


OtherProceedings of the Annual AI Systems in Government Conference
CityWashington, DC, USA

ASJC Scopus subject areas

  • General Engineering


Dive into the research topics of 'Deep domain models for discourse analysis.'. Together they form a unique fingerprint.

Cite this