Supervised word sense disambiguation (WSD) for truly polysemous words (in
contrast to homonyms) is difficult for machine learning, mainly due to two
problems: the lack of sense-tagged training data and the sparsity of the matrix
of observed instances vs. features. At the same time, high accuracy is necessary
for WSD to be beneficial for high-level applications, such as information
retrieval, question answering, and machine translation. This work addresses
the above two problems through combining rich linguistic knowledge
and machine learning methods. First, it proposes and demonstrates empirically
evidence that careful design and generation of linguistically motivated
features help to alleviate the data sparseness inherent in WSD. A state-of-theart
supervised system for verb sense disambiguation was introduced. Exploration
in three specific aspects of feature generation was discussed and
shown to elevate the system accuracy to top-level. It also shows the effectiveness
of active learning in the creation of more labeled training data for supervised
WSD - reducing the required training data by 1/2 to 3/4 when learning
coarse-grained English verb senses. The book is addressed to researchers in
Computer and Information Science and Computational Linguistics.