Watset: Local-Global Graph Clustering with Applications in Sense and Frame Induction

TL;DR

Watset is a versatile fuzzy graph clustering algorithm that disambiguates nodes to improve clustering accuracy, demonstrating competitive results in semantic and linguistic tasks.

Contribution

The paper introduces Watset, a novel meta-algorithm for fuzzy graph clustering that effectively handles node ambiguity and applies broadly across linguistic domains.

Findings

Competitive results in synset induction from synonymy graphs

Effective in semantic frame induction from dependency triples

Applicable to various linguistic network analyses

Abstract

We present a detailed theoretical and computational analysis of the Watset meta-algorithm for fuzzy graph clustering, which has been found to be widely applicable in a variety of domains. This algorithm creates an intermediate representation of the input graph that reflects the "ambiguity" of its nodes. Then, it uses hard clustering to discover clusters in this "disambiguated" intermediate graph. After outlining the approach and analyzing its computational complexity, we demonstrate that Watset shows competitive results in three applications: unsupervised synset induction from a synonymy graph, unsupervised semantic frame induction from dependency triples, and unsupervised semantic class induction from a distributional thesaurus. Our algorithm is generic and can be also applied to other networks of linguistic data.