SP-10K: A Large-scale Evaluation Set for Selectional Preference Acquisition

TL;DR

This paper introduces SP-10K, a large-scale evaluation dataset for selectional preference models, enabling better assessment and understanding of SP's role in language understanding and commonsense reasoning.

Contribution

The paper presents SP-10K, a comprehensive dataset for evaluating SP models, and demonstrates its utility in correlating SP with commonsense knowledge and coreference resolution.

Findings

SP-10K covers 10,000 SP pairs across five relations.

Evaluation of three SP acquisition methods shows their effectiveness.

SP-10K correlates with ConceptNet5, indicating its relevance to commonsense knowledge.

Abstract

Selectional Preference (SP) is a commonly observed language phenomenon and proved to be useful in many natural language processing tasks. To provide a better evaluation method for SP models, we introduce SP-10K, a large-scale evaluation set that provides human ratings for the plausibility of 10,000 SP pairs over five SP relations, covering 2,500 most frequent verbs, nouns, and adjectives in American English. Three representative SP acquisition methods based on pseudo-disambiguation are evaluated with SP-10K. To demonstrate the importance of our dataset, we investigate the relationship between SP-10K and the commonsense knowledge in ConceptNet5 and show the potential of using SP to represent the commonsense knowledge. We also use the Winograd Schema Challenge to prove that the proposed new SP relations are essential for the hard pronoun coreference resolution problem.