Computing Pronoun Antecedents in an English Query System

Abstract

In this paper I discuss how the Datalog English query system resolves pronominal references to extra-sentential antecedents that represent database records. When the system encounters a pronoun in a query, it searches through saved representations of earlier queries for an antecedent. A number of criteria must be satisfied before a proposed antecedent will be accepted. Among these are satisfaction of the pronoun's grammatical features and tests for contradictions and tautologies. Additional discriminators are applied in the event that there are two competing antecedents being considered. Of special interest is use of a hold queue mechanism which allows relaxation of the grammatical features of number and gender expressed by a personal pronoun. All of the strategies are independent of any application domain and do not fall into those parts of the system that need to be replaced or modified to interface Datalog to a new database.