Knowledge lean word sense disambiguation

摘要: We present a corpus-based approach to word-sense disambiguation that only requires information can be automatically extracted from untagged text. use unsupervised techniques estimate the parameters of model describing conditional distribution sense group given known contextual features. Both EM algorithm and Gibbs Sampling are evaluated determine which is most appropriate for our data. compare their accuracy in an experiment with thirteen different words three feature sets. results small but consistent improvement over algorithm.