SENS: the System for Evaluating Noun Sequences

Lucy Vanderwende

Abstract

This chapter discusses how to determine the relation that holds between nouns in a noun sequence. For natural language understanding, a system must be able to discover, for example, that there is a Locative relation in “vegetable market” but not in “lemon peel,” which contains, rather, a Part-of relation. A system is described here that will determine these relations automatically. First a classification of possible relations is presented, formulated as wh-questions. Since there is only a limited number of these relations that can possibly hold in a noun sequence, the system considers each possible relation in turn and provides an indication of how likely the relation is between two nouns by evaluating the semantic properties of each of the two nouns as evidenced by information extracted from online dictionaries. The result is the most likely interpretation of the noun sequence, together with the most likely senses of each noun.

Details

Publication typeInbook
URLhttp://rd.springer.com/chapter/10.1007/978-1-4615-3170-8_13
ChapterChapter 13, Editors Karen Jensen, George E. Heidorn, Stephen D. Richardson
PublisherKluwer International Series in Engineering and Computer Science
> Publications > SENS: the System for Evaluating Noun Sequences