Creator: Mihalcea, Rada
Description: This paper discusses a semi-complete disambiguation algorithm for open text. Word Sense Disambiguation (WSD) is one of the most difficult areas of Natural Language Processing (NLP); the semantic comprehension of a text, and the possibility to expand a text with semantically related information, drastically depends on the availability of a highly accurate WSD algorithm. Solutions considered so far by researchers for the WSD problem, are making use of machine readable dictionaries (Leacock, Chodorow and Miller 1998), or the information gathered from raw or semantically disambiguated corpora (Yarowsky 1995). These methods are designed either to work with a few pre-selected words, in which case a high accuracy is obtained, or they are general methods which disambiguate, with lower precision, all the words in a text. With the present work, the authors are trying to achieve a compromise between these two different directions. There are fields in NLP, like Information Retrieval and others, which could benefit from a method which performs a semi-complete disambiguation (i.e. it disambiguates only a certain percentage of the words in a text), but which is highly accurate.
Contributing Partner: UNT College of Engineering