Natural Language Agreement in the Generation Mechanism based on Stratified Graphs

Abstract

A Stratified Graph can generate natural language constructions by acting as a transition network with arc labels referring to word categories and word forms. In this representation, each grammar rule can be transposed in a sequence of labeled paths. The generation mechanism defined by means of stratified graph uses a bottom-up approach. The advantage of this generation mechanism relies on its bottom-up approach characteristic: this mechanism never suggests constructions that are not correct based on the involved representation elements. As it will be presented, implementing syntactic agreement rules in this natural language generation mechanism can determine different results even if identical words are considered in the generation mechanism.