<![CDATA[his study aims to develop a discourse structure-based deep learning model to improve the coherence and cohesion of students' expository texts. The study used the Research and Development (R&D) method, adapting the development steps according to Borg and Gall, simplified into six stages: needs analysis, model design, prototype development, expert validation, limited trials, and product revision. The research subjects consisted of three expert validators and thirty 10th-grade high school students. The research instruments included a needs analysis questionnaire, expert validation sheet, coherence-cohesion assessment rubric, and student response questionnaire. The results showed that the developed model received a very valid category from the experts with a total score of 88.5%. The validators assessed that the model met the appropriateness of the content, the completeness of the learning components, and the suitability of deep learning theory for its application in learning to write expository texts. From a practical aspect, the results of the limited trials showed that students could use the learning model easily, without technical or conceptual obstacles. From an effectiveness aspect, the results of the trials showed an increase in students' expository text writing ability scores, especially in the aspects of coherence and cohesion. Students demonstrated improved ability to connect paragraphs, use conjunctions appropriately, and construct cause-and-effect relationships logically. This improvement demonstrates that the discourse structure-based deep learning model is capable of encouraging the in-depth thinking processes required in expository writing. Therefore, this learning model is deemed feasible, practical, and effective for improving the quality of students' expository texts, particularly in developing coherent, clear, and organized discourse structures.]]>
Copyrights © 2025
