A hybrid model for arabic automatic question generation on transformer model with textrank
Not available
| 第一著者: | |
|---|---|
| その他の著者: | |
| フォーマット: | Master's thesis |
| 言語: | 英語 |
| 出版事項: |
Universiti Teknologi Malaysia
2025
|
| 主題: | |
| オンライン･アクセス: | https://utmik.utm.my/handle/123456789/39880 |
| Abstract | Abstract here |
| _version_ | 1854975068541550592 |
|---|---|
| author | Al-hashedi, Saleh Saleh Hussein |
| author2 | Norhaida Mohd. Suaib, supervisor |
| author_facet | Norhaida Mohd. Suaib, supervisor Al-hashedi, Saleh Saleh Hussein |
| author_sort | Al-hashedi, Saleh Saleh Hussein |
| description | Not available |
| format | Master's thesis |
| id | utm-123456789-39880 |
| institution | Universiti Teknologi Malaysia |
| language | English |
| publishDate | 2025 |
| publisher | Universiti Teknologi Malaysia |
| record_format | dspace |
| record_pdf | Abstract |
| spelling | utm-123456789-398802025-08-20T23:44:20Z A hybrid model for arabic automatic question generation on transformer model with textrank Al-hashedi, Saleh Saleh Hussein Norhaida Mohd. Suaib, supervisor Computer engineeering Not available Questions play a crucial role in education, enhancing learning outcomes for students of all ages. Automatic Question Generation (AQG) involves generating relevant questions from text data, potentially with accompanying answers. However, AQG in Arabic has seen limited progress, relying on manually constructed question styles using rule-based methods and input text from children's books or textbooks. These models suffer from restricted linguistic diversity, posing challenges when dealing with more extensive texts. This study addressed this gap by introducing SUALL, an end-to-end Arabic Automatic Question Generation (AAQG) hybrid model. The term "SUALL" originates from the Arabic word " ??? ? " meaning "Question" in English, symbolizing the model's ability to generate multiple interrogative questions from a single, unlimited-length document. Leveraging on the highly successful adaptable deep-learning Transformer encoder-decoder BERT-base architecture widely applied in various natural language processing (NLP) tasks, SUALL incorporates the TextRank algorithm, thus, overcoming the BERT's token limit of 512 tokens. To achieve this, the document was read and divided into paragraphs, and the TextRank algorithm was applied to extract key sentences from each paragraph. These sentences were then fed into the base model to generate questions. Notably, the SUALL model achieved promising results on the Microsoft Machine Reading Comprehension (mMARCO) dataset, showcasing a substantial 19.12% improvement in Bilingual Evaluation Understudy (BLEU), 23.00% in Metric for Evaluation of Translation with Explicit ORdering (METEOR), and 51.99% in Recall-Oriented Understudy for Gisting Evaluation (ROUGE-L) scores. This research significantly contributes to AAQG by addressing the limited research in Arabic, offering linguistic diversity and scalability to handle extensive texts. SUALL's capacity to generate multiple questions is a valuable support for educators and learners across diverse educational resources. zulraizam UTM 165 p. Thesis (Master of Philosophy (Computer Science)) - Universiti Teknologi Malaysia, 2023 2025-03-05T05:39:09Z 2025-03-05T05:39:09Z 2023-02 Master's thesis https://utmik.utm.my/handle/123456789/39880 vital:156456 valet-20240320-095915 ENG Closed Access UTM Complete Unpublished Completion application/pdf Universiti Teknologi Malaysia |
| spellingShingle | Computer engineeering Al-hashedi, Saleh Saleh Hussein A hybrid model for arabic automatic question generation on transformer model with textrank |
| thesis_level | Master |
| title | A hybrid model for arabic automatic question generation on transformer model with textrank |
| title_full | A hybrid model for arabic automatic question generation on transformer model with textrank |
| title_fullStr | A hybrid model for arabic automatic question generation on transformer model with textrank |
| title_full_unstemmed | A hybrid model for arabic automatic question generation on transformer model with textrank |
| title_short | A hybrid model for arabic automatic question generation on transformer model with textrank |
| title_sort | hybrid model for arabic automatic question generation on transformer model with textrank |
| topic | Computer engineeering |
| url | https://utmik.utm.my/handle/123456789/39880 |
| work_keys_str_mv | AT alhashedisalehsalehhussein ahybridmodelforarabicautomaticquestiongenerationontransformermodelwithtextrank AT alhashedisalehsalehhussein hybridmodelforarabicautomaticquestiongenerationontransformermodelwithtextrank |