Unsupervised Domain Tuning to Improve Word Sense Disambiguation

The topic of a document can prove to be useful information for Word Sense Disambiguation (WSD) since certain meanings tend to be associated with particular topics. This paper presents an LDA-based approach for WSD, which is trained using any available WSD system to establish a sense per (Latent Dirichlet allocation based) topic. The technique is tested using three unsupervised and one supervised WSD algorithms within the SPORT and FINANCE domains giving a performance increase each time, suggesting that the technique may be useful to improve the performance of any available WSD system.