--- title: "Enhancing Arabic Dialectal Sentiment Analysis through" description: "The paper tackles sentiment classification of hotel and tourism reviews written in Arabic dialects. 5 references, cited by 1 paper here." authors: ["Md. Rafiul Biswas", "Wajdi Zaghouani"] year: 2025 venue: "Proceedings of the Shared Task on Sentiment Analysis for Arabic Dialects" source: "https://aclanthology.org/2025.ranlp-ahasis.4/" licence: "CC BY 4.0 (ACL Anthology)" licence_url: "https://creativecommons.org/licenses/by/4.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/biswas-2025-enhancing-arabic-dialectal-sentiment-analysis/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/biswas-2025-enhancing-arabic-dialectal-sentiment-analysis/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2025](https://md.research.elamur.ai/free/en/papers/2025/page/14/) / Enhancing Arabic Dialectal Sentiment Analysis through Advanced Data Augmentation Techniques # Enhancing Arabic Dialectal Sentiment Analysis through Advanced Data Augmentation Techniques **Authors:** [Md. Rafiul Biswas](https://md.research.elamur.ai/free/en/authors/md-rafiul-biswas/) and [Wajdi Zaghouani](https://md.research.elamur.ai/free/en/authors/wajdi-zaghouani/) · **Year:** 2025 · **Venue:** [Proceedings of the Shared Task on Sentiment Analysis for Arabic Dialects](https://md.research.elamur.ai/free/en/venues/ranlp/) ([RANLP 2025](https://md.research.elamur.ai/free/en/venues/ranlp/2025/)) · **Publisher:** [INCOMA Ltd.](https://md.research.elamur.ai/free/en/publishers/incoma-ltd/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/) **Links:** [Publisher's page](https://aclanthology.org/2025.ranlp-ahasis.4/) · [PDF](https://aclanthology.org/2025.ranlp-ahasis.4.pdf) · [OpenAlex](https://openalex.org/W7140086310) **Citations:** 0 citations (OpenAlex, 2026-10-14) · [1](https://md.research.elamur.ai/free/en/papers/biswas-2025-enhancing-arabic-dialectal-sentiment-analysis/cited-by/) in our library, of the 149,439 papers whose bibliographies we read ## Summary The paper tackles sentiment classification of hotel and tourism reviews written in Arabic dialects. Training data for this setting is very scarce. The authors took part in the Ahasis shared task, which supplies only 860 labelled training sentences split evenly between Moroccan Darija and Saudi Arabic. To enlarge this set, they built a three-part augmentation pipeline: - paraphrasing with an LLM through the Fanar API, - generating new sentences from patterns mined in the originals, - swapping words for hospitality-specific synonyms. Each part tries to keep the dialect, the domain and the sentiment label intact. A fallback step switches to another method when the preferred one breaks, and humans checked the outputs. AraBERT-Large-v02 was then fine-tuned on the combined original and synthetic data. It scored 0.78 on validation and 0.75 on test across accuracy, precision, recall and F1. The shared-task winner reached an F1 of 0.81. The work shows that LLM-driven, label-aware augmentation is a practical way to stretch a tiny dialectal corpus. However, the paper gives no numeric score for a model trained without augmentation, so the size of the gain is not measured directly. ## Key points - **Data:** The Ahasis training set has 430 Darija and 430 Saudi sentences. Both halves have the same label mix: roughly 39% negative, 25% neutral and 36% positive. The test set holds 108 sentences per dialect. - **Augmented set sizes, Darija / Saudi:** - negative: 647 / 617 - positive: 613 / 587 - neutral: 421 / 394 - **LLM prompting:** The Fanar prompt contained a system and a user instruction plus ten worked examples. Each example listed the text, its dialect and its sentiment. - **Preprocessing:** HTML was stripped, and URLs, emails and mentions were replaced with placeholder tokens. Diacritics (tashkeel) and tatweel were removed. Emojis were deliberately kept as sentiment signals. - **Training setup:** - model: AraBERT-Large-v02, with CAMeLBERT mix also mentioned - split: stratified 80/20 train/validation - optimiser: AdamW at 2e-5 - other settings: 10 epochs, batch size 16, maximum length of 128 tokens, gradient clipping at 1.0 - validation loss levelled off around epoch 6 ## Interesting ideas - The LLM prompt is conditioned on both the dialect and the gold label, with labelled few-shot examples, so each synthetic sentence arrives already annotated and needs no relabelling. - The augmenters are chained with automatic fallback: if free generation fails, the system rebuilds a sentence from previously extracted structural patterns instead of dropping the item. - Word substitution is kept minimal and limited to synonyms from the target domain, which lowers the chance that a swap silently flips the label. - Augmentation quality is judged on three separate axes: meaning preservation, unchanged label distribution, and linguistic soundness. - Normalisation is aggressive for noise such as diacritics, elongation and URLs, but emojis are preserved because they carry affect. ## For Mova How these ideas could serve Mova, our small open language model — its files, stages and tests: - **I1** (learn): LLM teacher tiers could paraphrase treebank or fable sentences while being told which frame roles must survive. Witness parsers would then accept only paraphrases whose agreed trees keep the same who-did-what-to-whom. - **I2** (back): generating text from the meaning graph could try fluent generation first and fall back to deterministic templates tied to the graph. This way every graph still yields explainable text. - **I3** (world): WordNet synsets restricted to one ontology branch could produce minimal-pair variants of test stories. These would check that the world model's entities and states stay the same under lexical swaps. - **I4** (learn): silver-data batches could be rejected when their label or construction distribution drifts from the gold source. This would sit beside negative controls as a cheap automatic sanity gate. - **I5** (tokens): the tokenizer could keep emojis as their own tagged tokens rather than discarding them. The prag crate could then use them as cues for speech acts and irony. *Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.* ## Institutions - [Hamad bin Khalifa University](https://md.research.elamur.ai/free/en/universities/hamad-bin-khalifa-university/) — University, Doha, Qatar · 141 works · 14,175 citations - [Northwestern University in Qatar](https://md.research.elamur.ai/free/en/universities/northwestern-university-in-qatar/) — University, Doha, Qatar · 24 works · 148 citations 📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works 🧭 Topics (2): [😊 Sentiment & emotion](https://md.research.elamur.ai/free/en/topics/sentiment-emotion/) · 972 works · [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 2,908 works ## Cite Cite as: Md. Rafiul Biswas and Wajdi Zaghouani. 2025. Enhancing Arabic Dialectal Sentiment Analysis through Advanced Data Augmentation Techniques. Proceedings of the Shared Task on Sentiment Analysis for Arabic Dialects. The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{biswas2025enhancing, title = {{Enhancing Arabic Dialectal Sentiment Analysis through Advanced Data Augmentation Techniques}}, author = {Md. Rafiul Biswas and Wajdi Zaghouani}, booktitle = {Proceedings of the Shared Task on Sentiment Analysis for Arabic Dialects}, year = {2025}, doi = {10.26615/978-954-452-109-7-004}, url = {https://doi.org/10.26615/978-954-452-109-7-004} } ``` - [← previous English Pronunciation Evaluation without Complex Joint Training: LoRA Fine-tuned Speech Multimodal LLM](https://md.research.elamur.ai/free/en/papers/ahn-2025-english-pronunciation-evaluation-complex-joint/) - [next → Enhancing Arabic NLP Tasks through Character-Level Models and Data Augmentation](https://md.research.elamur.ai/free/en/papers/mohamed-2025-enhancing-arabic-nlp-tasks-through/) - [↑ back to papers of 2025](https://md.research.elamur.ai/free/en/papers/2025/page/14/) ## References 5 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. SA Salloum. 2021. Sentiment analysis in dialectal arabic: a systematic review. Advanced machine learning technologies and applications: proceedings of AMLTA. [find](https://scholar.google.com/scholar?q=SA+Salloum%2E+2021%2E+Sentiment+analysis+in+dialectal+arabic%3A+a+systematic+review%2E+Advanced+machine+learning+technologies+and+applications%3A+proceedings+of+AMLTA%2E) 2. Abdulmohsen Al-Thubaity, Qubayl Alqahtani, and Abdulaziz Aljandal. 2018. Sentiment lexicon for sentiment analysis of saudi dialect tweets. Procedia computer science, 142:301–307. [find](https://scholar.google.com/scholar?q=Abdulmohsen+Al%2DThubaity%2C+Qubayl+Alqahtani%2C+and+Abdulaziz+Aljandal%2E+2018%2E+Sentiment+lexicon+for+sentiment+analysis+of+saudi+dialect+tweets%2E+Procedia+computer+science%2C+142%3A301%E2%80%93307%2E) 3. Maram Alharbi, Salmane Chafik, Saad Ezzini, Ruslan Mitkov, Tharindu Ranasinghe, and Hansi Hettiarachchi. 2025a. [Ahasis: Shared task on sentiment analysis for arabic dialects](https://md.research.elamur.ai/free/en/papers/alharbi-2025-ahasis-shared-task-sentiment-analysis/). In Proceedings of the 15th International Conference on Recent Advances in Natural Language Processing (RANLP). 4. Uzair Shah, Md Rafiul Biswas, Marco Agus, Mowafa Househ, and Wajdi Zaghouani. 2024. [Mememind at araieval shared task: generative augmentation and feature fusion for multimodal propaganda detection in arabic memes through advanced language and vision models](https://md.research.elamur.ai/free/en/papers/shah-2024-mememind-araieval-shared-task-generative/). In Proceedings of the Second Arabic Natural Language Processing Conference, pages 467– 472. 5. Maram Alharbi, Saad Ezzini, Hansi Hettiarachchi, Tharindu Ranasinghe, and Ruslan Mitkov. 2025b. Evaluating large language models on arabic dialect Fanar Team, Ummar Abbas, Mohammad Shahmeer Ahmad, Firoj Alam, Enes Altinisik, Ehsannedin Asgari, Yazan Boshmaf, Sabri Boughorbel, Sanjay Chawla, Shammur Chowdhury, Fahim Dalvi, Kareem Darwish, Nadir Durrani, Mohamed Elfeky, Ahmed Elmagarmid, Mohamed Eltabakh, Masoomali Fatehkia, Anastasios Fragkopoulos, Maram Hasanain, Majd Hawasly, Mus’ab Husaini, Soon-Gyo Jung, Ji Kim Lucas, Walid Magdy, Safa Messaoud, Abubakr Mohamed, Tasnim Mohiuddin, Basel Mousi, Hamdy Mubarak, Ahmad Musleh, Zan Naeem, Mourad Ouzzani, Dorde Popovic, Amin Sadeghi, Husrev Taha Sencar, Mohammed Shinoy, Omar Sinan, Yifan Zhang, Ahmed Ali, Yassine El Kheir, Xiaosong Ma, and Chaoyi Ruan. 2025. [Fanar: An arabic-centric multimodal generative ai platform](https://md.research.elamur.ai/free/en/papers/abbas-2025-fanar-arabic-centric-multimodal-generative/). *Updated 31 October 2026*