Analyzing Language in Restricted Domains
Constraining the Interpretation of Nominal Compounds in a Limited Context
January 1, 1986
Nominal compounds are notoriously ambiguous. Syntactically, the number of parse trees of a nominal compound is exponential in the number of terms making up the compound. Worse, the semantic relationship between each term and its modifier has been deleted from the surface form and must be inferred. This makes nominal compounds even more highly ambiguous from a semantic point of view. The seemingly inherent ambiguity of nominal compounds might suggest that their utility is limited. On the contrary, they are very heavily used, especially in the context of a sublanguage. In earlier work, we described a system that synthesized candidate semantic interpretations of nominal compounds from possible interpretations of their constituents. The candidates were partially ordered with respect to an appropriateness measure based on a number of syntactic, semantic and pragmatic factors. More recent work has examined the additional constraints that arise in the context of an extended dialogue. In addition, we discuss the use of nominal compounding as a short-term naming device in these two situations.