Natural language text processing and the maximal join operator

Abstract

Conceptual graphs have proven to be a convenient representation formalism for natural language processing. In NLP, the lexical semantics of words often has to be mapped onto large knowledge bases. The most useful operation in the conceptual graphs formalism is the maximal join. This paper describes problems with ambiguities at different levels, including ambiguities resulting from the maximal join operation itself. This research was motivated by the idea to use different sources of information for text understanding. Layout information is one such source of semantic constraints. The traditional maximal join algorithm does not take into account any other constraints. In the following paper, we describe an implementation of an augmented maximal join operation. We will talk about the types of constraints and how they are integrated into the maximal join operation.