Skip to main content Skip to main navigation

Publikation

Domain Specific Sense Disambiguation with Unsupervised Methods

Diana Steffen; Bogdan Sacaleanu; Paul Buitelaar
In: Claudia Kunze; Lothar Lemnitzer; Andreas Wagner (Hrsg.). Anwendungen des deutschen Wortnetzes in Theorie und Praxis. Tagungsband des 1. GermaNet-Workshops des GLDV-AK Lexikografie. GermaNet-Workshops des GLDV-Arbeitskreises "Lexikografie", October 9-10, Tübingen, Germany, Pages 79-86, ISBN 3-9809183-0-0, Seminar für Sprachwissenschaft, Abt. Computerlinguistik, Tübingen, 10/2003.

Zusammenfassung

Most approaches in sense disambiguation have been restricted to supervised training over manually annotated, non-technical, English corpora. Application to a new language or technical domain requires extensive manual annotation of appropriate training corpora. As this is both expensive and inefficient, unsupervised methods are to be preferred, specifically in technical domains such as medicine. In the context of a project in the medical domain, we developed and evaluated two unsupervised methods for sense disambiguation.