| アイテムタイプ |
会議発表論文 / Conference Paper(1) |
| 公開日 |
2026-02-19 |
| タイトル |
|
|
タイトル |
ARxHYOKA at TAQEEM2025: Comparative Approaches to Arabic Essay Trait Scoring |
| 言語 |
|
|
言語 |
eng |
| 資源タイプ |
|
|
資源タイプ |
conference paper |
| アクセス権 |
|
|
アクセス権 |
open access |
| 著者 |
Alnajjar, Mohamad
Almoustafa, Ahmad
西山, 智弘
若宮, 翔子
荒牧, 英治
Matsuzaki, Takuya
|
| 抄録 |
|
|
内容記述タイプ |
Abstract |
|
内容記述 |
Arabic automated essay scoring (AES) presents unique challenges due to the linguistic complexity of Arabic and the need for rubric-specific evaluation. In this paper, we present ARxHYOKA, our submission to TAQEEM2025 Task B, which targets trait-specific AES using the Core Academic Skills Test (CAST) rubric. We evaluate four approaches: (1) GPT-based few-shot prompting, (2) fine-tuning BERTbased models, (3) classical machine learning approaches with embeddings and handcrafted features, and (4) fine-tuning text-generation large language models (LLMs). Our bestperforming system, GPT-4.1 with 10-shot CoT prompting, achieved the highest official score, outperforming all other approaches in average Quadratic Weighted Kappa (QWK) in the test phase. Fine-tuned BERT-based models performed on par with both the shared-task baseline and our GPT prompting setup in the development phase, while classical machine learning methods trailed these systems, and the finetuned Arabic LLM ranked last. We provide comparative analyses across systems to inform future research on Arabic AES. |
| 書誌情報 |
en : Proceedings of The Third Arabic Natural Language Processing Conference: Shared Tasks
p. 977-982,
ページ数 6,
発行日 2025
|
| 会議情報 |
|
|
|
会議名 |
ArabicNLP 2025 |
|
|
開始年 |
2025 |
|
|
開始月 |
11 |
|
|
開始日 |
08 |
|
|
終了年 |
2025 |
|
|
終了月 |
11 |
|
|
終了日 |
09 |
|
|
開催期間 |
2025-11-08 - 2025-11-09 |
|
|
開催地 |
Suzhou, China |
|
開催国 |
CHN |
| 出版者 |
|
|
出版者 |
Association for Computational Linguistics |
| 出版者版DOI |
|
|
関連タイプ |
isReplacedBy |
|
|
識別子タイプ |
DOI |
|
|
関連識別子 |
https://doi.org/10.18653/v1/2025.arabicnlp-sharedtasks.135 |
| 出版者版URI |
|
|
関連タイプ |
isReplacedBy |
|
|
識別子タイプ |
URI |
|
|
関連識別子 |
https://aclanthology.org/2025.arabicnlp-sharedtasks.135/ |
| 権利 |
|
|
権利情報Resource |
https://creativecommons.org/licenses/by/4.0/ |
|
権利情報 |
©2025 Association for Computational Linguistics. Materials published in or after 2016 are licensed on a Creative Commons Attribution 4.0 International License. |
| 著者版フラグ |
|
|
出版タイプ |
NA |
| 助成情報 |
|
|
|
助成機関名 |
National Center for Global Health and Medicine (NCGM) |
|
|
研究課題番号 |
JPJ012425 |
|
|
研究課題名 |
Cross-ministerial Strategic Innovation Promotion Program (SIP) |
