Train-O-Matic: Large-Scale Supervised Word Sense Disambiguation in Multiple Languages without Manual Training Data

摘要

Annotating large numbers of sentencesudwith senses is the heaviest requirementudof current Word Sense Disambiguation.udWe present Train-O-Matic, a languageindependentudmethod for generating millionsudof sense-annotated training instancesudfor virtually all meanings of words inuda language’s vocabulary. The approachudis fully automatic: no human interventionudis required and the only type of humanudknowledge used is a WordNet-likeudresource. Train-O-Matic achieves consistentlyudstate-of-the-art performance acrossudgold standard datasets and languages,udwhile at the same time removing the burdenudof manual annotation. All the traininguddata is available for research purposes atudhttp://trainomatic.org