A Generic Syntax Analyzer for the Computerization of African Languages

Abstract

This article enters within the framework of a contribution to the computerization of African Languages. It presents a generic parsing tool designed for West African languages despite their individual specificities. The genericity of the tool resides in the fact that it is adaptable to any West African language while using a minimum of resources. It only needs an etiquette dictionary and context-free grammar. From these two resources, the tool is designed to generate an LR (1) parser. The choice of this technique is based on its supremacy over other techniques despite its disadvantage of generating an exponentially large parsing table. However, to reduce the size of the automaton while maintaining the power of the technique, the system brings together dictionary entries depending on their type, gender, and number in the form of terminals. The results obtained show that it is possible to generate a relatively small and efficient analysis table instead of a large one. An application of the proposed tool to the Hausa nominal phrases has succeeded in reducing the number of terminals from more than 10,000 to only 88. Instead of having an automaton with thousands of states, we end up with an automaton with only 198 states.