Project Abstract:

Word sense disambiguation refers to the task of determining the correct meaning of a word in context. For example, the word “plant” has multiple meanings including “factory” and “flora” depending on the context in which it is used. Word sense disambiguation can generally be done well by learning machines that have been trained using supervised learning methods.

Unfortunately, to train a word sense disambiguation model with a coverage of about 3000 words using a standard supervised learning method would require an estimated 16 man-years of human annotation effort. This project explores the possibility of reducing the amount of human annotation required for word sense disambiguation through the use of semi-supervised machine learning methods that reduce the need for human annotation by exploiting the large amount of unannotated documents that are easily available.