Mierzwa, The Formal Model of Computer Decomposition of Natural Language Sentences into Groups of Words for Computer Inference Purposes (Abstract)

One of the aims of NLP is to inference with knowledge included in natural language sentences. To enable such a process, a suitable representation of sentences in computer memory is required. One of the representations, for which an inference process was worked out, is a symbolic expression representation of natural language sentences. The main element of this formalism is splitting sentences into groups of words. The work presents a formal method of splitting sentences into groups of words for inference purposes with symbolic expression representation. The method uses for grammar representation the Structural Link Grammars formalism--SLG introduced by the author (based on the Link Grammar). For SLG, the next elements were worked out: (1) assigning a set of links for each word (based on lexical categories and list of features), (2) a method of building a graphical representation of processed sentences, (3) an algorithm of splitting sentences into groups of words based on obtained graphical representation, and (4) a realization of English grammar in SLG (as an example). This method combines some features of dependency grammar and phrase structure grammar enabling to write relations between words and groups into symbolic expression.

Link Grammar Bibliography page

Daniel Sleator