Graph-based unsupervised learning of word similarities using heterogeneous feature types

Abstract

In this work, we propose a graph-based approach to computing similarities between words in an unsupervised manner, and take advantage of heterogeneous feature types in the process. The approach is based on the creation of two separate graphs, one for words and one for features of different types (alignment-based, orthographic, etc.). The graphs are connected through edges that link nodes in the feature graph to nodes in the word graph, the edge weights representing the importance of a particular feature for a particular word. High quality graphs are learned during training, and the proposed method outperforms experimental baselines.