The syntactic and semantic processing of english comparatives
Maria L. Gini, Michael B. Kac, Duane Olawsky · University of Minnesota eBooks · 1990
Comparative expressions (CE's) such as bigger than and oranges than are highly ambiguous, and their meaning is context dependent. Furthermore, CE's are structurally complex. Thus, they pose problems for the semantic and syntactic processing algorithms typically used in natural language database interfaces. We address both the syntactic and semantic problems. In treating the semantics of CE's, we focus on polysemous comparison attribute ambiguities (i.e. determining what quantities should be used to compare two entities). To resolve these ambiguities our natural language interface interacts with the user, finding out which of the possible interpretations was intended. We have extended a rule-based multi-level semantic processing algorithm to facilitate this interaction by recognizing the occurrence of comparison attribute ambiguity and then calculating and presenting a list of candidate comparison attributes from which the user may choose. To handle the syntactic complexity of CE's, we have designed and implemented a new type of multi-phase natural language parser, called Genefer, inspired by linguistic work in Corepresentational Grammar. This grammatical theory allows an elegant syntactic description of CE's. Our parser is markedly different from previous parsers since it follows the interdictive grammar enforcement paradigm. Genefer demonstrates that interdictive parsing is indeed a viable alternative to the more traditional approbatory methods. We have also developed techniques for obtaining an initial semantic representation from the output of Genefer.