Computational models of incremental semantic interpretation
Nicholas J. Haddock · Language and Cognitive Processes · 1989
This paper provides a tutorial introduction to the computational problem of “incremental” semantic interpretation: the problem of computing the semantic interpretation of a sentence on a word-by-word basis as the sentence is read from left to right. We also consider the way in which incremental interpretation can guide the syntactic processing, and pay particular attention to the use of referential information during parsing. We address these issues in terms of the computational models of Bobrow and Webber (1980a, b), Winograd (1972), Mellish (1985), and Haddock (1988). In each case, we ask two questions: First, to what extent does the program embody an incremental process of semantic interpretation? Secondly, how does this process help resolve syntactic ambiguities encountered by the parser? Here we will relate the behaviour of the models to the psychological theory of ambiguity resolution offered by Crain and Steedman (1985). In the context of this review, we identify some of the specific problems involved in incremental interpretation, and compare the computational techniques which have been applied to these problems.