An Analogical Parser for Restricted Domains

1. T H E P R O B L E M W I T H P A R S E R S A parser is a device that provides a description of the syntactic phrases that make up a sentence. For a speech understanding task such as ATIS, the parser has two roles. First, it should provide a description of the phrases in a sentence so these phrases can be interpreted by a subsequent semantic processor. The second function is to provide a language model a model of the likelihood of a sentence to constrain the speech recognition task. It is unfortunately the case that existing parsers developed for text fulfill neither of these roles very well. It is useful to begin by reviewing some of the reasons for this failure. We can describe the situation in terms of three general problems that parsers face: the Lexicality Problem, the Tail Problem, and the Interpolation Problem. The Lexicality Problem The most familiar way to think of a parser is as a device that provides a description of a sentence given some grammar. Consider for example a context free grammar, where nonterminal categories are rewritten as terminals or nonterminals, and terminals are rewritten as words. There typically is no way to express the constraints among individual words. Yet it is clear that much of our knowledge of language has to do with what words go together.[2] Merely knowing the grammatical rules of the language is not enough to predict which words can go together. So for example, general English grammatical rules admit premodification of a noun by another noun or by an adjective. It is possible to describe broad semantic constraints on such modification; so for example, early morning is a case of a time-adjective modifying a time-period, and morning flight is a time-period modifying an event. Already we are have an explosion of categories in the grammar, since we are talking not about nouns and adjectives, but about a fairly detailed subclassification of semantic types of nouns and adjectives. But the problem is worse than this. As Table 1 shows, even this rough characterization of semantic constraints on modification is insufficient, since the adjective-noun combination early night does not occur. This dependency of syntactic combinability on particular lexicM items is repeated across the grammar and lexicon. The lexicality problem has two aspects. One is representing the information and the other is acquiring it. There has recently been increasing work on both aspects of the problem. The approach described in this paper is but one of many possible approaches, designed with an emphasis on facilitating efficient parsing.