On dealing with quantification in natural language utterances

Abstract

This paper presents some preliminary and still quite rudimentary ideas growing out of on-going studies into the problem of representation of quantified information in English, in terms of linked and nested “attribute-value” structures. Specifically, we indicate how such representation can be accomplished within the kind of attributevalue framework provided by Heidorn's NLP (Natural Language Processor) system. A scheme for representing well-formed formulas of a sorted first-order logic (with descriptions) is first presented and serves as a bare conceptual framework for understanding the sequel. Then we indicate how certain commonly encountered forms of English language quantificational utterances may be reduced to canonical forms representable within such an underlying scheme. A few of many still outstanding logico-linguistic issues (viz.: ambiguities connected with ‘a’, ‘any’ and ‘some’; multiple predication; and anonymous quantification) are noted and discussed. Numerous examples are included. © 1978, Academic Press Inc. (London) Limited. All rights reserved.