A Survey of Techniques for Unsupervised Word Sense Induction

Many applications in natural language processing benefit from the use of word senses rather than surface word forms. While the use of word senses has historically required large, manually compiled dictionaries, recent work has focused on automatically inducing these senses from unannotated text. This paper presents an overview of the task of unsupervised word sense induction (WSI) and compares several approaches to the task, concluding with a final overview of the techniques surveyed.