Linguistic Rules Based Approach for Automatic Restoration of Accents on French Texts

Abstract

Nowadays, in the context of email as well as many other domains, there are more and more French texts wrongly accented or completely unaccented. Furthermore, it should be noted that in French, the accent has a value and a linguistic function. It expresses the language’s subtleties and especially allows avoiding ambiguities and misinterpretation. Even though in most cases the loss of information resulting from the absence of accents is not a major issue for human beings, it is very problematic for automatic processing of text and increases the ambiguity involved in Natural Language Processing. However, it gets tedious to do this manually hence the importance of automatic accent restoration systems. In this perspective, this paper aims at presenting a novel system for the automatic restoration of accents in French texts. Unlike a few existing approaches using statistical methods, our approach is essentially based on linguistic rules that are more reliable.