Knowledge Sources for Word Sense Disambiguation

TL;DR

This paper analyzes the relationship between knowledge types and information sources used in Word Sense Disambiguation, comparing various algorithms to guide future system development and knowledge acquisition strategies.

Contribution

It systematizes the connection between desired knowledge types and actual information sources in WSD, and compares algorithm performances on a common test setting.

Findings

Comparison of algorithms on a unified test set

Insights into the effectiveness of different knowledge sources

Guidance for shifting from information-based to knowledge-based systems

Abstract

Two kinds of systems have been defined during the long history of WSD: principled systems that define which knowledge types are useful for WSD, and robust systems that use the information sources at hand, such as, dictionaries, light-weight ontologies or hand-tagged corpora. This paper tries to systematize the relation between desired knowledge types and actual information sources. We also compare the results for a wide range of algorithms that have been evaluated on a common test setting in our research group. We hope that this analysis will help change the shift from systems based on information sources to systems based on knowledge sources. This study might also shed some light on semi-automatic acquisition of desired knowledge types from existing resources.