---
title: "NeuralNoodles@CHiPSAL 2026 — paper by Muntaha et al. (2026)"
description: "The CUET team describes its entry to Subtask B of the CHiPSAL 2026 shared task, which asks for a… 19 references, cited by 2 papers here."
authors: ["Sidratul Muntaha", "Sabila Anzum", "Arpita Mallik", "Hasan Murad"]
year: 2026
venue: "Proceedings of the Second workshop on Challenges in Processing South Asian Languages (CHiPSAL2026)"
source: "https://aclanthology.org/2026.chipsal-1.25/"
doi: "10.63317/5om43abhz5vy"
licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)"
licence_url: "https://creativecommons.org/licenses/by-nc/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/muntaha-2026-neuralnoodles-chipsal-2026-late-fusion/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/muntaha-2026-neuralnoodles-chipsal-2026-late-fusion/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/8/) / NeuralNoodles@CHiPSAL 2026: Late-Fusion Multimodal Stacking for Nepali Meme Sentiment Classification

# NeuralNoodles@CHiPSAL 2026: Late-Fusion Multimodal Stacking for Nepali Meme Sentiment Classification

**Authors:** Sidratul Muntaha, Sabila Anzum, [Arpita Mallik](https://md.research.elamur.ai/free/en/authors/arpita-mallik/) and [Hasan Murad](https://md.research.elamur.ai/free/en/authors/hasan-murad/) · **Year:** 2026 · **Venue:** [Proceedings of the Second workshop on Challenges in Processing South Asian Languages (CHiPSAL2026)](https://md.research.elamur.ai/free/en/venues/chipsal/) ([workshop on Challenges in Processing South Asian Languages 2026](https://md.research.elamur.ai/free/en/venues/chipsal/2026/)) · **Publisher:** [European Language Resources Association (ELRA)](https://md.research.elamur.ai/free/en/publishers/european-language-resources-association-elra/) · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.chipsal-1.25/) · [PDF](https://aclanthology.org/2026.chipsal-1.25.pdf) · [DOI 10.63317/5om43abhz5vy](https://doi.org/10.63317/5om43abhz5vy) · [OpenAlex](https://openalex.org/W7166514838)

**Citations:** 0 citations (OpenAlex, 2026-10-14) · [2](https://md.research.elamur.ai/free/en/papers/muntaha-2026-neuralnoodles-chipsal-2026-late-fusion/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
Meme sentiment comes from how the image and its embedded caption work together. This is hard to model for Nepali, where labelled multimodal data are scarce. The CUET team describes its entry to Subtask B of the CHiPSAL 2026 shared task, which asks for a Negative, Neutral or Positive label for each Nepali text-embedded meme. The training set has only 1,061 memes, with Neutral as the largest class, and 133 memes are held out for testing.

The authors avoid a heavy cross-attention transformer. Instead they use late fusion with three independent base models:
- word- and character-level TF-IDF with Logistic Regression,
- frozen paraphrase-multilingual-MiniLM-L12-v2 sentence embeddings with Logistic Regression,
- a fine-tuned EfficientNet-B0 image classifier.

A meta-classifier combines their class probabilities through cross-validated stacking. The fused system scored Macro F1 0.4234 in cross-validation, beating the unimodal and pairwise combinations. On the official test set it scored 0.5045 and placed 6th. The work offers a lightweight, reproducible, open-source baseline for multimodal sentiment analysis in low-resource languages.

## Key points
- **Data:** CHiPSAL 2026 Subtask B data, with 247 Positive, 341 Negative and 473 Neutral training memes, plus a 133-meme test set. Text comes from OCR, and Macro F1 is the evaluation metric.
- **Text features:**
  - TF-IDF combines word 1–2-grams (up to 3,000 features) and character 3–5-grams (up to 2,000 features).
  - The 384-dimensional MiniLM embeddings are kept frozen to limit overfitting.
- **Image model:** EfficientNet-B0 with ImageNet weights, fine-tuned with AdamW (learning rate 1e-4, batch size 16) for 4 epochs per fold in stratified 5-fold CV. Images are augmented with flips and ±10° rotations.
- **Results:**
  - The image-only model was the strongest single modality.
  - Two-modality pairs did not beat it, scoring 0.3715–0.3897.
  - Fold scores for the stacked model ranged from 0.4072 to 0.4548.
- **Per-class F1:** Negative 0.5026, Neutral 0.4338, Positive 0.3339. Positive memes were often mislabelled as Negative, which the authors attribute partly to sarcasm. Code is on GitHub (NeuralNoodles-CHiPSAL-2026).

*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

- [Chittagong University of Engineering & Technology](https://md.research.elamur.ai/free/en/universities/chittagong-university-of-engineering-technology/) — University, Chittagong, Bangladesh · 25 works · 29 citations

📚 Shelves (3): [🖥️ 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 · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 2,869 works

🧭 Topics (2): [😊 Sentiment & emotion](https://md.research.elamur.ai/free/en/topics/sentiment-emotion/) · 972 works · [🖼️ Vision & language](https://md.research.elamur.ai/free/en/topics/multimodal/) · 657 works

## Cite

Cite as: Sidratul Muntaha, Sabila Anzum, Arpita Mallik and Hasan Murad. 2026. NeuralNoodles@CHiPSAL 2026: Late-Fusion Multimodal Stacking for Nepali Meme Sentiment Classification. Proceedings of the Second workshop on Challenges in Processing South Asian Languages (CHiPSAL2026).

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

```bibtex
@inproceedings{muntaha2026neuralnoodles,
  title = {{NeuralNoodles@CHiPSAL 2026: Late-Fusion Multimodal Stacking for Nepali Meme Sentiment Classification}},
  author = {Sidratul Muntaha, Sabila Anzum, Arpita Mallik and Hasan Murad},
  booktitle = {Proceedings of the Second workshop on Challenges in Processing South Asian Languages (CHiPSAL2026)},
  year = {2026},
  doi = {10.63317/5om43abhz5vy},
  url = {https://doi.org/10.63317/5om43abhz5vy}
}
```

- [← previous Neural network embeddings recover value dimensions from psychometric survey items on par with human data](https://md.research.elamur.ai/free/en/papers/pellert-2026-neural-network-embeddings-recover-value/)
- [next → NSF-SciFy: Mining the NSF Awards Database for Scientific Claims](https://md.research.elamur.ai/free/en/papers/rao-2026-nsf-scify-mining-nsf-awards/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/8/)

## References

19 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Aashish Bhandari, Siddhant B Shah, Surendrabikram Thapa, Usman Naseem, and Mehwish Nasim. 2023. Crisishatemm: Multimodal analysis of directed and undirected hate speech in text-embedded images from russia-ukraine conflict. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Reco… [find](https://scholar.google.com/scholar?q=Aashish+Bhandari%2C+Siddhant+B+Shah%2C+Surendrabikram+Thapa%2C+Usman+Naseem%2C+and+Mehwish+Nasim%2E+2023%2E+Crisishatemm%3A+Multimodal+analysis+of+directed+and+undirected+hate+speech+in+text%2Dembedded+images+from+ru)
2. Ana-Maria Bucur, Adrian Cosma, and Ioan-Bogdan Iordache. 2022. Blue at memotion 2.0 2022: You have my image, my text and my transformer. [find](https://scholar.google.com/scholar?q=Ana%2DMaria+Bucur%2C+Adrian+Cosma%2C+and+Ioan%2DBogdan+Iordache%2E+2022%2E+Blue+at+memotion+2%2E0+2022%3A+You+have+my+image%2C+my+text+and+my+transformer%2E)
3. Kazi Toufique Elahi, Tasnuva Binte Rahman, Shakil Shahriar, Samir Sarker, Sajib Kumar Saha Joy, and Faisal Muhammad Shah. 2023. Explainable multimodal sentiment analysis on bengali memes. [find](https://scholar.google.com/scholar?q=Kazi+Toufique+Elahi%2C+Tasnuva+Binte+Rahman%2C+Shakil+Shahriar%2C+Samir+Sarker%2C+Sajib+Kumar+Saha+Joy%2C+and+Faisal+Muhammad+Shah%2E+2023%2E+Explainable+multimodal+sentiment+analysis+on+bengali+memes%2E)
4. A. Guleria, K. Varshney, G. Pahwa, et al. 2025. Multimodal sentiment analysis of english and hinglish memes. Multimedia Tools and Applications, 84:15331–15356. [find](https://scholar.google.com/scholar?q=A%2E+Guleria%2C+K%2E+Varshney%2C+G%2E+Pahwa%2C+et+al%2E+2025%2E+Multimodal+sentiment+analysis+of+english+and+hinglish+memes%2E+Multimedia+Tools+and+Applications%2C+84%3A15331%E2%80%9315356%2E)
5. Md Hasan, Nusratul Jannat, Eftekhar Hossain, Omar Sharif, and Mohammed Moshiul Hoque. 2022. CUET-NLP@DravidianLangTech-ACL2022: Investigating deep learning techniques to detect multimodal troll memes. In Proceedings of the Second Workshop on Speech and Language Technologies for Dravidian Languages,… [find](https://scholar.google.com/scholar?q=Md+Hasan%2C+Nusratul+Jannat%2C+Eftekhar+Hossain%2C+Omar+Sharif%2C+and+Mohammed+Moshiul+Hoque%2E+2022%2E+CUET%2DNLP%40DravidianLangTech%2DACL2022%3A+Investigating+deep+learning+techniques+to+detect+multimodal+troll+memes%2E)
6. Eftekhar Hossain, Omar Sharif, and Mohammed Moshiul Hoque. 2022. MemoSen: A multimodal dataset for sentiment analysis of memes. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 1542–1554, Marseille, France. European Language Resources Association. [find](https://scholar.google.com/scholar?q=Eftekhar+Hossain%2C+Omar+Sharif%2C+and+Mohammed+Moshiul+Hoque%2E+2022%2E+MemoSen%3A+A+multimodal+dataset+for+sentiment+analysis+of+memes%2E+In+Proceedings+of+the+Thirteenth+Language+Resources+and+Evaluation+Confe)
7. Rishav Jha, Mohit Ranjan Panda, Shubham K C, and Aaditya Dahal. 2025. Deep learning architectures for multimodal sentiment analysis. In 2025 International Conference on Intelligent and Cloud Computing (ICoICC), pages 1–6. [find](https://scholar.google.com/scholar?q=Rishav+Jha%2C+Mohit+Ranjan+Panda%2C+Shubham+K+C%2C+and+Aaditya+Dahal%2E+2025%2E+Deep+learning+architectures+for+multimodal+sentiment+analysis%2E+In+2025+International+Conference+on+Intelligent+and+Cloud+Computing)
8. Nicola Messina, Fabrizio Falchi, Claudio Gennaro, and Giuseppe Amato. 2021. AIMH at SemEval-2021 task 6: Multimodal classification using an ensemble of transformer models. In Proceedings of the 15th International Workshop on Semantic Evaluation (SemEval-2021), pages 1020–1026, Online. Association fo… [find](https://scholar.google.com/scholar?q=Nicola+Messina%2C+Fabrizio+Falchi%2C+Claudio+Gennaro%2C+and+Giuseppe+Amato%2E+2021%2E+AIMH+at+SemEval%2D2021+task+6%3A+Multimodal+classification+using+an+ensemble+of+transformer+models%2E+In+Proceedings+of+the+15th+I)
9. Aishwaria Roy, Hasan Murad, and Ayman Iktidar. 2024. A multimodal approach to bangla cyberbullying meme detection in social media. In 2024 27th International Conference on Computer and Information Technology (ICCIT), pages 1886– 1891. [find](https://scholar.google.com/scholar?q=Aishwaria+Roy%2C+Hasan+Murad%2C+and+Ayman+Iktidar%2E+2024%2E+A+multimodal+approach+to+bangla+cyberbullying+meme+detection+in+social+media%2E+In+2024+27th+International+Conference+on+Computer+and+Information+Tec)
10. Kengatharaiyer Sarveswaran, Surendrabikram Thapa, Ashwini Vaidya, Tafseer Ahmed Khan, and Bal Krishna Bal. 2026. [Findings of the second workshop on challenges in processing south asian languages (chipsal 2026)](https://md.research.elamur.ai/free/en/papers/sarveswaran-2026-findings-second-workshop-challenges-processing/). In Proceedings of the Second Workshop on Challenges in Processing South Asian Languages (…
11. Kogilavani Shanmugavadivel, Malliga Subramanian, Mohamed Arsath H, Ramya K, and Ragav R. 2025. TEAM_STRIKERS@DravidianLangTech2025: Misogyny meme detection in Tamil using multimodal deep learning. In Proceedings of the Fifth Workshop on Speech, Vision, and Language Technologies for Dravidian Languag… [find](https://scholar.google.com/scholar?q=Kogilavani+Shanmugavadivel%2C+Malliga+Subramanian%2C+Mohamed+Arsath+H%2C+Ramya+K%2C+and+Ragav+R%2E+2025%2E+TEAM%5FSTRIKERS%40DravidianLangTech2025%3A+Misogyny+meme+detection+in+Tamil+using+multimodal+deep+learning%2E+In+)
12. Mayukh Sharma, Ilanthenral Kandasamy, and W.b. Vasantha. 2020. Memebusters at SemEval-2020 task 8: Feature fusion model for sentiment analysis on memes using transfer learning. In Proceedings of the Fourteenth Workshop on Semantic Evaluation, pages 1163–1171, Barcelona (online). International Commit… [find](https://scholar.google.com/scholar?q=Mayukh+Sharma%2C+Ilanthenral+Kandasamy%2C+and+W%2Eb%2E+Vasantha%2E+2020%2E+Memebusters+at+SemEval%2D2020+task+8%3A+Feature+fusion+model+for+sentiment+analysis+on+memes+using+transfer+learning%2E+In+Proceedings+of+the+F)
13. Shivam Sharma, Ramaneswaran S, Md. Shad Akhtar, and Tanmoy Chakraborty. 2024. Emotion-aware multimodal fusion for meme emotion detection. IEEE Transactions on Affective Computing, 15(3):1800–1811. [find](https://scholar.google.com/scholar?q=Shivam+Sharma%2C+Ramaneswaran+S%2C+Md%2E+Shad+Akhtar%2C+and+Tanmoy+Chakraborty%2E+2024%2E+Emotion%2Daware+multimodal+fusion+for+meme+emotion+detection%2E+IEEE+Transactions+on+Affective+Computing%2C+15%283%29%3A1800%E2%80%931811%2E)
14. Steve Durairaj Swamy, Shubham Laddha, Basil Abdussalam, Debayan Datta, and Anupam Jamatia. 2020. NIT-agartala-NLP-team at SemEval-2020 task 8: Building multimodal classifiers to tackle Internet humor. In Proceedings of the Fourteenth Workshop on Semantic Evaluation, pages 1179– 1189, Barcelona (onli… [find](https://scholar.google.com/scholar?q=Steve+Durairaj+Swamy%2C+Shubham+Laddha%2C+Basil+Abdussalam%2C+Debayan+Datta%2C+and+Anupam+Jamatia%2E+2020%2E+NIT%2Dagartala%2DNLP%2Dteam+at+SemEval%2D2020+task+8%3A+Building+multimodal+classifiers+to+tackle+Internet+humor%2E)
15. Surendrabikram Thapa, Shuvam Shiwakoti, Siddhant Bikram Shah, Kritesh Rauniyar, Laxmi Thapa, Surabhi Adhikari, Kristina T. Johnson, Kengatharaiyer Sarveswaran, Bal Krishna Bal, and Usman Naseem. 2026. [Multimodal hate and sentiment understanding in low-resource text-embedded images for online safety…](https://md.research.elamur.ai/free/en/papers/thapa-2026-multimodal-hate-sentiment-understanding-low/)
16. Surendrabikram Thapa, Hariram Veeramani, Liang Hu, Qi Zhang, Wei Wang, and Usman Naseem. 2025a. A multimodal prompt-based framework for analyzing code-mixed and low-resource memes. In Proceedings of the International AAAI Conference on Web and Social Media, volume 19, pages 1913–1923. [find](https://scholar.google.com/scholar?q=Surendrabikram+Thapa%2C+Hariram+Veeramani%2C+Liang+Hu%2C+Qi+Zhang%2C+Wei+Wang%2C+and+Usman+Naseem%2E+2025a%2E+A+multimodal+prompt%2Dbased+framework+for+analyzing+code%2Dmixed+and+low%2Dresource+memes%2E+In+Proceedings+of+t)
17. Surendrabikram Thapa, Hariram Veeramani, Imran Razzak, Roy Ka-Wei Lee, and Usman Naseem. 2025b. Cross platform multimodal retrieval augmented distillation for code-switched content understanding. In Companion Proceedings of the ACM on Web Conference 2025, pages 2042– 2051. [find](https://scholar.google.com/scholar?q=Surendrabikram+Thapa%2C+Hariram+Veeramani%2C+Imran+Razzak%2C+Roy+Ka%2DWei+Lee%2C+and+Usman+Naseem%2E+2025b%2E+Cross+platform+multimodal+retrieval+augmented+distillation+for+code%2Dswitched+content+understanding%2E+In+C)
18. Georgi Vankov, Dimitar Dimitrov, Ivan Koychev, and Preslav Nakov. 2025. Multimodal sentiment analysis: Recognizing sentiment in memes. In Artificial Intelligence: Methodology, Systems, and Applications, pages 1–11, Cham. Springer Nature Switzerland. [find](https://scholar.google.com/scholar?q=Georgi+Vankov%2C+Dimitar+Dimitrov%2C+Ivan+Koychev%2C+and+Preslav+Nakov%2E+2025%2E+Multimodal+sentiment+analysis%3A+Recognizing+sentiment+in+memes%2E+In+Artificial+Intelligence%3A+Methodology%2C+Systems%2C+and+Application)
19. Riza Velioglu and Jewgeni Rose. 2020. Detecting hate speech in memes using multimodal deep learning approaches: Prize-winning solution to hateful memes challenge. [find](https://scholar.google.com/scholar?q=Riza+Velioglu+and+Jewgeni+Rose%2E+2020%2E+Detecting+hate+speech+in+memes+using+multimodal+deep+learning+approaches%3A+Prize%2Dwinning+solution+to+hateful+memes+challenge%2E)

*Updated 31 October 2026*