---
title: "Assessing the Performance of ChatGPT-4, Fine-tuned BERT and"
description: "The authors present this as the first evaluation of an LLM on MA sentiment. 35 references, cited by 2 papers here."
authors: ["Mohamed Hannani", "Abdelhadi Soudi", "Kristof Van Laerhoven"]
year: 2024
venue: "Proceedings of the 4th International Conference on Natural Language Processing for Digital Humanities"
source: "https://aclanthology.org/2024.nlp4dh-1.47/"
doi: "10.18653/v1/2024.nlp4dh-1.47"
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/hannani-2024-assessing-performance-chatgpt-4-fine/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/hannani-2024-assessing-performance-chatgpt-4-fine/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2024](https://md.research.elamur.ai/free/en/papers/2024/page/10/) / Assessing the Performance of ChatGPT-4, Fine-tuned BERT and Traditional ML Models on Moroccan Arabic Sentiment Analysis

# Assessing the Performance of ChatGPT-4, Fine-tuned BERT and Traditional ML Models on Moroccan Arabic Sentiment Analysis

**Authors:** Mohamed Hannani, [Abdelhadi Soudi](https://md.research.elamur.ai/free/en/authors/abdelhadi-soudi/) and Kristof Van Laerhoven · **Year:** 2024 · **Venue:** [Proceedings of the 4th International Conference on Natural Language Processing for Digital Humanities](https://md.research.elamur.ai/free/en/venues/nlp4dh/) ([International Conference on Natural Language Processing for the Digital Humanities 2024](https://md.research.elamur.ai/free/en/venues/nlp4dh/2024/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2024.nlp4dh-1.47/) · [PDF](https://aclanthology.org/2024.nlp4dh-1.47.pdf) · [DOI 10.18653/v1/2024.nlp4dh-1.47](https://doi.org/10.18653/v1/2024.nlp4dh-1.47) · [OpenAlex](https://openalex.org/W4404780537)

**Citations:** 4 citations (OpenAlex, 2026-10-14) · [2](https://md.research.elamur.ai/free/en/papers/hannani-2024-assessing-performance-chatgpt-4-fine/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
The paper asks how well current sentiment classifiers handle Moroccan Arabic (MA, Darija). MA is a low-resource dialect that mixes regional forms, loanwords, and both Arabic and Latin script. The authors compare four families of systems on two public corpora:

- **gpt-4-turbo**, used zero-shot through LangChain.
- **Several BERT-family encoders**, either fully fine-tuned or with a frozen backbone, each topped by a small custom classifier head.
- **FastText embeddings**.
- **Classical ML classifiers**.

The corpora are MAC, a set of tweets labelled positive/negative/neutral, and MYC, a set of YouTube comments labelled positive/negative and written in both scripts. The authors present this as the first evaluation of an LLM on MA sentiment. It matters because Arabic sentiment work has mostly covered MSA in Arabic script, while real Moroccan online text is dialectal and often Latin-scripted. The extracted text we received leaves out the results section. This summary therefore reports no scores or model ranking; consult the full paper for those.

## Key points
- **Datasets:**
  - MAC: about 18k manually labelled tweets, with an associated lexicon of 30,000 polarity-tagged words. After the "mixed" class and unlabelled rows were dropped, it has 9,888 positive, 3,505 negative and 4,039 neutral items (12,145 tagged MSA, 5,287 MA).
  - MYC: about 20k raw comments from 50 popular Moroccan YouTube channels, reduced to about 16k after cleaning (7,427 positive, 8,621 negative).
- **Encoders compared:** bert-base-multilingual-cased, bert-base-arabic, darijabert-arabizi, DarijaBERT and bert-base-arabertv2, with vocabularies from 32k to about 120k.
- **LLM setup:**
  - gpt-4-turbo with temperature 0.3 and top_p 1.
  - The prompt is a LangChain SystemMessage plus HumanMessage that asks for a numeric label only (0/1/2 for MAC, 0/2 for MYC).
  - The reply goes through JsonOutputToolsParser to enforce that format.
- **Classifier head:** the BERT pooler output (768 dimensions) feeds a Linear layer to 512, then ReLU, Dropout and a Softmax over 3 (MAC) or 2 (MYC) classes.
- **Data audit:** the authors found that the released "pre-processed" MYC was not actually cleaned. They re-applied the steps described in its paper: dropping empty comments, usernames, http/https links and unlabelled rows.

## Interesting ideas
- Wrapping a zero-shot LLM call with a strict output parser that accepts only a bare class code turns a chatty model into a dependable labeller. Malformed answers can then be caught instead of silently mis-scored.
- Running the same pretrained encoders both fully fine-tuned and with a frozen backbone separates what the pretrained representation already knows from what task training adds.
- Instead of trusting a dataset's claim that it was cleaned, the authors re-ran the documented preprocessing and found the release did not match. This is a cheap audit that protects every downstream number.
- Tagging each item with its language variety or script (MSA vs MA, Arabic vs Latin) lets evaluation be broken down by the hardest slices rather than reported only as one average.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (learn): LLM teacher tiers could demand a fixed JSON schema for each label they supply and reject anything else, so teacher disagreements reflect real ambiguity rather than formatting noise.
- **I2** (tree): gparse could be trained both ways on the same treebank to see how much parsing skill comes from the encoder itself and whether full fine-tuning is worth its cost.
- **I3** (learn): before any treebank or silver corpus enters training, a checker could re-apply its stated filters and conversion rules and flag any mismatch between claimed and actual state.
- **I4** (tokens): test sets for the en tokenizer and tagger could carry tags such as informal spelling, social-media style or non-standard orthography, so accuracy is reported per slice and weak spots are visible.

*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

- [National School of Mineral Industry](https://md.research.elamur.ai/free/en/universities/national-school-of-mineral-industry/) — University, Rabat, Morocco · 4 works · 114 citations

Also on the byline: [University of Siegen](https://ror.org/02azyry73).

📚 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: Mohamed Hannani, Abdelhadi Soudi and Kristof Van Laerhoven. 2024. Assessing the Performance of ChatGPT-4, Fine-tuned BERT and Traditional ML Models on Moroccan Arabic Sentiment Analysis. Proceedings of the 4th International Conference on Natural Language Processing for Digital Humanities.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{hannani2024assessing,
  title = {{Assessing the Performance of ChatGPT-4, Fine-tuned BERT and Traditional ML Models on Moroccan Arabic Sentiment Analysis}},
  author = {Mohamed Hannani, Abdelhadi Soudi and Kristof {Van Laerhoven}},
  booktitle = {Proceedings of the 4th International Conference on Natural Language Processing for Digital Humanities},
  year = {2024},
  doi = {10.18653/v1/2024.nlp4dh-1.47},
  url = {https://doi.org/10.18653/v1/2024.nlp4dh-1.47}
}
```

- [← previous ASOS at KSAA-CAD 2024: One Embedding is All You Need for Your Dictionary](https://md.research.elamur.ai/free/en/papers/sibaee-2024-asos-ksaa-cad-2024-one/)
- [next → Capturing the full potential of Maltese language learning through ChatGPT](https://md.research.elamur.ai/free/en/papers/zammit-2024-capturing-full-potential-maltese-language/)
- [↑ back to papers of 2024](https://md.research.elamur.ai/free/en/papers/2024/page/10/)

## References

35 references, 20 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. The low recall values in MYC dataset indicate that the traditional classifiers have difficulty in identifying all instances of the positive class. In other words, they tend to miss a significant number of positive samples. Potential reasons for the low recall could include differences in data distribution, domain-specific characteristics, or noise introduced during data collection. Additionally, the language References or dialectal variations present in MYC data, distinct Muhammad Abdul-Mageed, Mona Diab, and Sandra from those in MAC, might pose challenges for clas-Kübler. 2014. [Samar: Subjectivity and sentiment sifiers in accurately identifying positive instances](https://md.research.elamur.ai/free/en/papers/abdul-mageed-2021-arbert-marbert-deep-bidirectional-transformers/). analysis for arabic social media. Computer Speech Prior work by Jbel et al. (2023) laid the ground-& Language, 28(1):20–37. work for sentiment analysis on the MYC dataset Muhammad Abdul-Mageed, AbdelRahim Elmadany, by creating the dataset and evaluating a range of and El Moatez Billah Nagoudi. 2020. Arbert & traditional and neural network models. They re-marbert: Deep bidirectional transformers for arabic. ported that the best performance was achieved with arXiv preprint arXiv:2101.01785. 496 [↗](https://arxiv.org/abs/2101.01785)
2. Nawaf A Abdulla, Nizar A Ahmed, Mohammed A Shehab, and Mahmoud Al-Ayyoub. 2013. Arabic sentiment analysis: Lexicon-based and corpus-based. In 2013 IEEE Jordan conference on applied electrical engineering and computing technologies (AEECT), pages 1–6. IEEE. [find](https://scholar.google.com/scholar?q=Nawaf+A+Abdulla%2C+Nizar+A+Ahmed%2C+Mohammed+A+Shehab%2C+and+Mahmoud+Al%2DAyyoub%2E+2013%2E+Arabic+sentiment+analysis%3A+Lexicon%2Dbased+and+corpus%2Dbased%2E+In+2013+IEEE+Jordan+conference+on+applied+electrical+engineer)
3. Ahmed Y Al-Obaidi and Venus W Samawi. 2016. Opinion mining: Analysis of comments written in arabic colloquial. In Proceedings of the World Congress on Engineering and Computer Science, volume 1. [find](https://scholar.google.com/scholar?q=Ahmed+Y+Al%2DObaidi+and+Venus+W+Samawi%2E+2016%2E+Opinion+mining%3A+Analysis+of+comments+written+in+arabic+colloquial%2E+In+Proceedings+of+the+World+Congress+on+Engineering+and+Computer+Science%2C+volume+1%2E)
4. Abdulmohsen Al-Thubaity, Sakhar Alkhereyf, Hanan Murayshid, Nouf Alshalawi, Maha Omirah, Raghad Alateeq, Rawabi Almutairi, Razan Alsuwailem, Manal Alhassoun, and Imaan Alkhanen. 2023. [Evaluating ChatGPT and bard AI on Arabic sentiment analysis](https://md.research.elamur.ai/free/en/papers/thubaity-2023-evaluating-chatgpt-bard-ai-arabic/). In Proceedings of ArabicNLP 2023, pages 335–349, Singapore (Hybrid). Association for Computational Linguistics.
5. Nora Al-Twairesh, Hend Al-Khalifa, and AbdulMalik Al-Salman. 2016. [Arasenti: Large-scale twitter-specific arabic sentiment lexicons](https://md.research.elamur.ai/free/en/papers/twairesh-2016-arasenti-large-scale-twitter-specific/). In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 697–705.
6. Bouamor, Karim Bouzoubaa, Violetta Cavalli-Sforza, Samhaa R El-Beltagy, Wassim El-Hajj, et al. 2021. A panoramic survey of natural language processing in the arab world. Communications of the ACM, 64(4):72–81. [find](https://scholar.google.com/scholar?q=Bouamor%2C+Karim+Bouzoubaa%2C+Violetta+Cavalli%2DSforza%2C+Samhaa+R+El%2DBeltagy%2C+Wassim+El%2DHajj%2C+et+al%2E+2021%2E+A+panoramic+survey+of+natural+language+processing+in+the+arab+world%2E+Communications+of+the+ACM%2C+64%28)
7. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. [Bert: Pre-training of deep bidirectional transformers for language understanding](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/). arXiv preprint arXiv:1810.04805. [↗](https://arxiv.org/abs/1810.04805)
8. Rehab M Duwairi and Islam Qarqaz. 2014. Arabic sentiment analysis using supervised classification. In 2014 International Conference on Future Internet of Things and Cloud, pages 579–583. IEEE. [find](https://scholar.google.com/scholar?q=Rehab+M+Duwairi+and+Islam+Qarqaz%2E+2014%2E+Arabic+sentiment+analysis+using+supervised+classification%2E+In+2014+International+Conference+on+Future+Internet+of+Things+and+Cloud%2C+pages+579%E2%80%93583%2E+IEEE%2E)
9. AbdelRahim Elmadany, Muhammad Abdul-Mageed, et al. 2022. [Arat5: Text-to-text transformers for arabic language generation](https://md.research.elamur.ai/free/en/papers/nagoudi-2022-arat5-text-text-transformers-arabic/). In Proceedings of the 60th annual meeting of the association for computational linguistics (Volume 1: Long papers), pages 628–647.
10. Hady ElSahar and Samhaa R El-Beltagy. 2014. A fully automated approach for arabic slang lexicon extraction from microblogs. In International conference on intelligent text processing and computational linguistics, pages 79–91. Springer. [find](https://scholar.google.com/scholar?q=Hady+ElSahar+and+Samhaa+R+El%2DBeltagy%2E+2014%2E+A+fully+automated+approach+for+arabic+slang+lexicon+extraction+from+microblogs%2E+In+International+conference+on+intelligent+text+processing+and+computational)
11. Mostafa M Amin, Erik Cambria, and Björn W Schuller. 2023. Will affective computing emerge from foundation models and general artificial intelligence? a first evaluation of chatgpt. IEEE Intelligent Systems, 38(2):15–23. [↗](https://doi.org/10.1109/mis.2023.3254179)
12. Moncef Garouani and Jamal Kharroubi. 2021. Mac: an open and free moroccan arabic corpus for sentiment analysis. In The Proceedings of the International Conference on Smart City Applications, pages 849– 858. Springer. [find](https://scholar.google.com/scholar?q=Moncef+Garouani+and+Jamal+Kharroubi%2E+2021%2E+Mac%3A+an+open+and+free+moroccan+arabic+corpus+for+sentiment+analysis%2E+In+The+Proceedings+of+the+International+Conference+on+Smart+City+Applications%2C+pages+849)
13. Rohan Anil, Andrew M Dai, Orhan Firat, Melvin Johnson, Dmitry Lepikhin, Alexandre Passos, Siamak Shakeri, Emanuel Taropa, Paige Bailey, Zhifeng Chen, et al. 2023. [Palm 2 technical report](https://md.research.elamur.ai/free/en/papers/anil-2023-palm-2-technical-report/). arXiv preprint arXiv:2305.10403. [↗](https://arxiv.org/abs/2305.10403)
14. Go Inoue, Bashar Alhafni, Nurpeiis Baimukan, Houda Bouamor, and Nizar Habash. 2021. [The interplay of variant, size, and task type in Arabic pre-trained language models](https://md.research.elamur.ai/free/en/papers/inoue-2021-interplay-variant-size-task-type/). In Proceedings of the Sixth Arabic Natural Language Processing Workshop, pages 92– 104, Kyiv, Ukraine (Virtual). Association for Computational Linguistics.
15. Wissam Antoun, Fady Baly, and Hazem Hajj. [Arabert: Transformer-based model for arabic language understanding](https://md.research.elamur.ai/free/en/papers/antoun-2020-arabert-transformer-based-model-arabic/). In LREC 2020 Workshop Language Resources and Evaluation Conference 11–16 May 2020, page 9. [↗](https://doi.org/10.48550/arxiv.2003.00104)
16. Adel Assiri, Ahmed Emam, and Hmood Al-Dossari. 2016. Saudi twitter corpus for sentiment analysis. International Journal of Computer and Information Engineering, 10(2):272–275. [↗](https://doi.org/10.5281/zenodo.1338815)
17. Ramy Baly, Alaa Khaddaj, Hazem Hajj, Wassim El-Hajj, and Khaled Bashir Shaban. 2019. [Arsentd-lev: A multi-topic corpus for target-based sentiment analysis in arabic levantine tweets](https://md.research.elamur.ai/free/en/papers/baly-2019-arsentd-lev-multi-topic-corpus/). arXiv preprint arXiv:1906.01830. [↗](https://arxiv.org/abs/1906.01830)
18. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. [Language models are few-shot learners](https://md.research.elamur.ai/free/en/papers/brown-2020-language-models-few-shot-learners/). Advances in neural information processing systems, 33:1877–1901.
19. Kareem Darwish, Nizar Habash, Mourad Abbas, Hend Al-Khalifa, Huseein T Al-Natsheh, Houda Mouad Jbel, Imad Hafidi, and Abdulmutallib Metrane. 2023. [Sentiment analysis dataset in moroccan dialect: Bridging the gap between arabic and latin scripted dialect](https://md.research.elamur.ai/free/en/papers/jbel-2023-sentiment-analysis-dataset-moroccan-dialect/). arXiv preprint arXiv:2303.15987. [↗](https://arxiv.org/abs/2303.15987)
20. Armand Joulin, Edouard Grave, Piotr Bojanowski, Matthijs Douze, H’erve J’egou, and Tomas Mikolov. 2016. [Fasttext.zip: Compressing text classification models](https://md.research.elamur.ai/free/en/papers/joulin-2016-fasttext-zip-compressing-text-classification/). arXiv preprint arXiv:1612.03651. [↗](https://arxiv.org/abs/1612.03651)
21. Armand Joulin, Edouard Grave, Piotr Bojanowski, and Tomas Mikolov. 2017. Bag of tricks for efficient text classification. In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers, pages 427–431. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Armand+Joulin%2C+Edouard+Grave%2C+Piotr+Bojanowski%2C+and+Tomas+Mikolov%2E+2017%2E+Bag+of+tricks+for+efficient+text+classification%2E+In+Proceedings+of+the+15th+Conference+of+the+European+Chapter+of+the+Associati)
22. Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, and Radu Soricut. 2019. [Albert: A lite bert for self-supervised learning of language representations](https://md.research.elamur.ai/free/en/papers/lan-2019-albert-lite-bert-self-supervised/). arXiv preprint arXiv:1909.11942. [↗](https://arxiv.org/abs/1909.11942)
23. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. [Roberta: A robustly optimized bert pretraining approach](https://md.research.elamur.ai/free/en/papers/liu-2019-roberta-robustly-optimized-bert-pretraining/). arXiv preprint arXiv:1907.11692. [↗](https://arxiv.org/abs/1907.11692)
24. Salima Mdhaffar, Fethi Bougares, Yannick Esteve, and Lamia Hadrich-Belguith. 2017. [Sentiment analysis of tunisian dialects: Linguistic ressources and experiments](https://md.research.elamur.ai/free/en/papers/medhaffar-2017-sentiment-analysis-tunisian-dialects-linguistic/). In Third Arabic natural language processing workshop (WANLP), pages 55–61. Xlnet: Generalized autoregressive pretraining for language understanding. Advances in neural information processing systems, 32.
25. Zhixue Zhao, George Chrysostomou, Kalina Bontcheva, and Nikolaos Aletras. 2022. On the impact of temporal concept drift on model explanations. In Findings of the Association for Computational Linguistics: EMNLP 2022, pages 4039–4054, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Zhixue+Zhao%2C+George+Chrysostomou%2C+Kalina+Bontcheva%2C+and+Nikolaos+Aletras%2E+2022%2E+On+the+impact+of+temporal+concept+drift+on+model+explanations%2E+In+Findings+of+the+Association+for+Computational+Linguist)
26. Mahmoud Nabil, Mohamed Aly, and Amir Atiya. 2015. [Astd: Arabic sentiment tweets dataset](https://md.research.elamur.ai/free/en/papers/nabil-2015-astd-arabic-sentiment-tweets-dataset/). In Proceedings of the 2015 conference on empirical methods in natural language processing, pages 2515–2519. [↗](https://doi.org/10.60692/3ee2g-zjt73)
27. OpenAI. 2023. [Gpt-4 technical report](https://md.research.elamur.ai/free/en/papers/openai-2023-gpt-4-technical-report/). abs/2303.08774. [↗](https://arxiv.org/abs/2303.08774)
28. ArXiv, Guilherme Penedo, Quentin Malartic, Daniel Hesslow, Ruxandra Cojocaru, Alessandro Cappelli, Hamza Alobeidli, Baptiste Pannier, Ebtesam Almazrouei, and Julien Launay. 2023. The refinedweb dataset for falcon llm: outperforming curated corpora with web data, and web data only. arXiv preprint arXiv:2306.01116. [↗](https://arxiv.org/abs/2306.01116)
29. Alec Radford, Karthik Narasimhan, Tim Salimans, Ilya Sutskever, et al. 2018. Improving language understanding by generative pre-training. [find](https://scholar.google.com/scholar?q=Alec+Radford%2C+Karthik+Narasimhan%2C+Tim+Salimans%2C+Ilya+Sutskever%2C+et+al%2E+2018%2E+Improving+language+understanding+by+generative+pre%2Dtraining%2E)
30. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. 2020. [Exploring the limits of transfer learning with a unified text-to-text transformer](https://md.research.elamur.ai/free/en/papers/raffel-2019-exploring-limits-transfer-learning-unified/). Journal of machine learning research, 21(140):1–67. [↗](https://doi.org/10.48550/arxiv.1910.10683)
31. Eshrag Refaee and Verena Rieser. 2014. [An arabic twitter corpus for subjectivity and sentiment analysis](https://md.research.elamur.ai/free/en/papers/refaee-2014-arabic-twitter-corpus-subjectivity-sentiment/). In 9th International Language Resources and Evaluation Conference, pages 2268–2273. European Language Resources Association.
32. Teven Le Scao, Angela Fan, Christopher Akiki, Ellie Pavlick, Suzana Ilić, Daniel Hesslow, Roman Castagné, Alexandra Sasha Luccioni, François Yvon, et al. 2022. Bloom: A 176b-parameter open-access multilingual language model. arXiv preprint arXiv:2211.05100. [↗](https://arxiv.org/abs/2211.05100)
33. Chi Sun, Xipeng Qiu, Yige Xu, and Xuanjing Huang. 2019. [How to fine-tune bert for text classification?](https://md.research.elamur.ai/free/en/papers/sun-2019-fine-tune-bert-text-classification/) In Chinese computational linguistics: 18th China national conference, CCL 2019, Kunming, China, October 18–20, 2019, proceedings 18, pages 194– 206. Springer. [↗](https://doi.org/10.48550/arxiv.1905.05583)
34. Zengzhi Wang, Qiming Xie, Zixiang Ding, Yi Feng, and Rui Xia. 2023. Is chatgpt a good sentiment analyzer? a preliminary study. ArXiv, abs/2304.04339. [↗](https://arxiv.org/abs/2304.04339)
35. Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Russ R Salakhutdinov, and Quoc V Le. 2019. [find](https://scholar.google.com/scholar?q=Zhilin+Yang%2C+Zihang+Dai%2C+Yiming+Yang%2C+Jaime+Carbonell%2C+Russ+R+Salakhutdinov%2C+and+Quoc+V+Le%2E+2019%2E)

*Updated 1 November 2026*