On The Current State Of Machine Translation: Investigating The Statistical Approach Limitations And The Neural Model Implications

Résumé: Translation is of increasing interest in our time due to the rapid expansion in the amount of content that should be transferred beyond the linguistic barriers of its mother tongue. Yet, human translation cannot keep up with the continuous flow of documents published today worldwide, which makes it imperative to use the available translation technologies, in particular machine translation tools. Nowadays, and thanks to the improvements made in the field of artificial intelligence, the statistical approach has been eventually replaced by the neural model. This paper attempted to explore this subject from theoretical and applied point of view in order to determine the motifs and the impact of this change on the quality of machine translation in the target language, by comparing the translation quality produced by ‘Reverso’ and ‘Google Translate’; to conclude that although neural machine translation gave promising results, the raw output is not comparable to a human translation.

Publié dans la revue: الاكاديمية للدراسات الاجتماعية و الانسانية