Coping with Syntactic Ambiguity or How to Put the Block in the Box on the Table

Abstract

Sentences are far more ambiguous than one might have thought. There may be hundreds, perhaps thousands of syntatic parse trees for certain very natural sentences of English. This fact has been a major problem confronting natural language processing because it indicates that it may require a long time to construct a list of all the parse trees, and furthermore, it isn't clear what to do with the list once it has ben constructed. This list may be so numerous that it is probably not the most convenient representation for communication with the semantic and pragmatic processing modules. In this paper we propose some methods for dealing with syntactic ambiguity in ways that take advantage of certain regularities among the alternative parse trees. These regularities will be expressed as linear combinations of ATN networks, and also as sums and products of formal power series. We will suggest some ways that practical processor can take advantage of this modularity in order to deal more efficiently with combinatoric ambiguity. In particular, we will show how a processor can efficiently compute the ambiguity of an input sentence (or any portion thereof). Furthermore, we will show how to compile certain grammars into a form that can be processed more efficiently. In some cases, including the "every way ambiguous" grammar (e.g., conjunction, prepositional phrases, noun-noun modification), processing time will be reduced from O9n^3) to O(n). Finally, we will show how to uncompile certain highly optimized grammars into a form suitable for linguistic analysis.