Journal of Studies in Language, Culture and Society (JSLCS)
Volume 9, Numéro 2, Pages 125-145
2026-06-30

Assessing Neural Machine Translation Errors And Post-editing Practices On Literary Traits Of Exhortation Texts



Authors : Zemni Bahia .

Abstract

Despite the challenges and limitations, it is undeniable that machine translation (MT) remains a crucial tool for translators, especially as technology developers endeavor to improve current systems and theorists strive to optimize concepts and processes, such as post-editing (PE). In this regard, the present study is dedicated to a critical and statistical analysis of the PE operations carried out by university students to improve the quality of an Arabic-English MT of a motivational text produced by the SYSTRAN Pro trial version. A total of 27 Master’s students specializing in translation participated in the study, and a mixed-method approach combining descriptive statistics and one-way ANOVA was employed to evaluate the data. The study focuses on categorizing and searching for statistically significant differences among the types of corrections proposed by students, and then analyzing them. The motivational text was chosen for its literary, religious, and cultural elements, which present particularly challenging translation problems for NMT systems. The results indicate limitations in participants’ proficiency in PE. The researchers noted that the participants presented four cases of corrections (either they identified the error without making any corrections (Case 1), or they suggested an inappropriate correction (Case 2), or they provided a correction that was partially accurate (Case 3), or they offered a relevant correction (Case 4)), which unveil significant statistical differences among these four correction cases. Notably, Case 4 exhibited a significantly higher frequency. These findings highlight the need for dedicated PE training within translation curricula to enhance students’ competence in revising NMT outputs.

Keywords

Neural Machine Translation (NMT) ; Post-editing ; SYSTRAN ; NMT quality ; Motivational text ; Arabic-to-English translation

Les articles similaires

Post-édition De Ta Neuronale à La Dgt Et Qualité Des Textes Finaux : étude De Cas     / Neural Machine Translation Post-editing In Dgt And Final Text Quality: A Case Study

De Faria Pires Loïc . 
pages 77-98.

Post-editing Machine Translation As A Vital Skill In The 21st Century: A Critical Analysis Of The Integration Of Artificial Intelligence In University-level Translation Education

Slimani Hafida .  Saim Rachid . 
pages 165-182.

The Importance Of Including Post-editing Of Machine Translation In The Curricula Of Translation Student, In Light Of Other’s Experiences

Bentahar Fares . 
pages 115-126.

Comparative Analysis Of Errors In Machine Translation And Computer-assisted Translation: Arabic–english Bidirectional Legal Translation And The Human Factor

Bougherara Antar .  Benaouda Adila . 
pages 344-359.

Exploration De La Traduction Automatique Neuronale Espagnol-français : Pour Une Traductologie De Corpus Appliquée à L’analyse Des Outils De Traduction Exploring Spanish-french Neural Machine Translation : Towards A Corpus-based Translation Studies Approach For The Analysis Of Translation Tools

Valdez Cristian .  Lomeña Galiano María . 
pages 86-112.

Formats de citation