Linguistic Error Analysis in ChatGPT’s Translation of Haqq at-Tilāwati by Husni Sheikh Osman

Abstract

The development of Artificial Intelligence (AI) has led to the increasing use of automatic translation for Arabic religious texts, including works on tajwid and qira’at. However, studies that specifically map the types of linguistic errors produced by machine translation in Arabic religious texts remain limited. This study aims to analyze linguistic errors in ChatGPT’s Arabic-Indonesian translations from morphological, syntactic, and semantic perspectives.This research employs a qualitative descriptive method with content analysis, using a corpus consisting of one chapter, Faslun fi at-Talfiq, from the book Ḥaqq at-Tilawati by Husni Syeikh Osman. The data were obtained from ChatGPT’s translations of the Arabic source text and were classified according to categories of linguistic errors.The findings reveal that out of 37 identified error instances, semantic errors are the most dominant with 28 data points (76%), followed by syntactic errors with 6 data points (16%) and morphological errors with 3 data points (8%). The dominance of semantic errors indicates ChatGPT’s limitations in handling domain-specific terminology in tajwid and qira’at studies. These findings provide an initial overview of ChatGPT’s linguistic error patterns in translating tajwid texts and may serve as a preliminary reference for the critical use of machine translation in religious texts.