A Comparison Between ChatGPT and DeepSeek in Fine Tuning a Bilingual Text

Document Type

Book Chapter

Source of Publication

Impacts of Capstone Projects for Future Ready Graduates Design Implementation and Evaluation

Publication Date

6-18-2026

Abstract

This study investigates which pretrained Transformer model, ChatGPT or DeepSeek, demonstrates greater potential for implementation in intelligent educational technologies that facilitate English-medium instruction alongside Arabic. A mixed-methods research approach was employed, amalgamating quantitative training metrics with qualitative evaluations obtained from six high-achieving Arabic-speaking undergraduate students. The results revealed that ChatGPT displayed strong convergence, heightened accuracy, and improved effectiveness in generating coherent, pedagogically aligned bilingual grammar content. Conversely, while DeepSeek produced coherent explanations in English, it exhibited significant shortcomings in its Arabic outputs, characterized by tense inaccuracies, grammatical inconsistencies, and dialectal mismatches. Participants reported higher satisfaction with ChatGPT for its clarity, cultural relevance, and alignment with the SIOP instructional components, whereas DeepSeek needed improvements in its Arabic modeling and bilingual integration.

ISBN

[9798337353128, 9798337353142]

Publisher

IGI Global Scientific Publishing

First Page

171

Last Page

192

Disciplines

Computer Sciences

Keywords

Grammar (0.68) | Computer science (0.61) | Natural language processing (0.57) | Artificial intelligence (0.51) | Linguistics (0.5) | Transformer (0.49) | Arabic (0.48) | Qualitative analysis (0.34) | Fine-tuning (0.3) | Psychology (0.3) | Past tense (0.3) | Computational linguistics (0.28) | Control (management) (0.27) | Qualitative research (0.27) | Protocol analysis (0.26) | Discourse analysis (0.26) | Applied linguistics (0.26)

Scopus ID

105043996145

Indexed in Scopus

yes

Open Access

no

Link to Full Text

Share

COinS