Finding Constraints for Semantic Relations via Clustering

Abstract

Automatic recognition of semantic relations constitutes an important part of information extraction. Many existing information extraction systems rely on syntactic information found in a sentence to accomplish this task. In this paper, we look into relation arguments and claim that some semantic relations can be described by constraints imposed on them. This information would provide more insight on the nature of semantic relations and could be further combined with the evidence found in a sentence to arrive at actual extractions.