A typology of ambiguity as it relates to natural language processing

Ambiguity and vagueness have always been major problems for natural language understanding systems because they are so pervasive in every natural language. People are often unaware of multiple senses, using such devices as world knowledge, lexical relationships and preferred readings to automatically select one reading. In a natural language processing system, these devices can be very difficult to build into the system. It is therefore important to have a clear understanding of the varieties of ambiguity and vagueness in order to decide how and whether to deal with each one of them computationally. This dissertation provides a typology of all of the possible forms of ambiguity and vagueness that impact natural language processing systems.; The typology was created by decomposing the possible types of ambiguity into a set of 26 binary features of ambiguity together with one level feature with possible values of morpheme, word, phrase, sentence or discourse. Then, the possible interactions of the features were examined to determine whether pairs of the features could both apply to the same ambiguity or not. Based on this examination, the full set of possible settings of the feature values was projected to create a typology of ambiguity and vagueness.; In order to establish the applicability of this typology to natural language processing, five representative semantic frameworks that have been used in such systems were selected. These are Dynamic Syntax, Discourse Representation Theory, Monotonic Syntax, Situation Semantics and Statistical methods. After an introduction to the problems of ambiguity and vagueness in natural language understanding, each of these semantic theories is briefly outlined, the features of ambiguity are presented, the feature interactions are determined, and each feature is examined as to how such an ambiguity would be handled within each semantic framework. Finally, a set of semantic tests is proposed to distinguish for each feature of ambiguity whether a given ambiguous or vague utterance is positive or negative for that feature. Examples of types of ambiguity are given throughout, and an appendix lists some naturally occurring instances of ambiguity and vagueness taken from newspapers, specified for the features of ambiguity.