The purpose of this review is to compare and contrast the literature on computational semantics relating to the representation of word-sense distinctions and the acquisition of lexical knowledge. This will support the research outlined above into automatically acquiring sense distinctions from on-line resources, such as machine-readable dictionaries and text corpora, and research into representing the distinctions using Bayesian networks. Therefore the review will also include some work on using Bayesian networks in natural language processing, as well as their use in general. The survey will mainly cover Computational Linguistics, but it will also include work in related fields such as Linguistics, Philosophy, and Psychology. In addition, work on statistical inferencing in various fields will be included, especially techniques related to Bayesian inferencing.
A common theme discussed throughout will be on how the approaches reviewed bear upon word-sense distinctions. For instance, Amsler [3] mainly discusses taxonomic development using machine readable resources, but he also provides an illustrative example of the features needed for capturing finer word-sense distinctions in a broad class of verbs. Another common theme will be the relation of the work to semantic relatedness, because differentia often can be used to determine semantic relatedness, either directly through functional attributes or indirectly through shared attributes. Richardson's thesis [101] demonstrates the benefits of doing so.