---
title: "CANVAS: A Multimodal Dataset of Chinese Textbook — Zhu 2026"
description: "Most bias benchmarks for Chinese (e.g., CHBias, CBBQ, McBE) target text, and vision fairness datasets such as FairFace or VISOGENDER deal with single…"
authors: ["Haotian Zhu", "Kefan Yu", "Min Li"]
year: 2026
venue: "Proceedings of the Fifteenth Language Resources and Evaluation Conference"
source: "https://aclanthology.org/2026.lrec-1.725/"
doi: "10.63317/2xmqax7numnm"
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/zhu-2026-canvas-multimodal-dataset-chinese-textbook/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/zhu-2026-canvas-multimodal-dataset-chinese-textbook/)

[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/4/) / CANVAS: A Multimodal Dataset of Chinese Textbook Images for Bias and Representation Analysis

# CANVAS: A Multimodal Dataset of Chinese Textbook Images for Bias and Representation Analysis

**Authors:** Haotian Zhu, Kefan Yu, Min Li · **Year:** 2026 · **Venue:** Proceedings of the Fifteenth Language Resources and Evaluation Conference · **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.lrec-1.725/) · [PDF](https://aclanthology.org/2026.lrec-1.725.pdf) · [DOI 10.63317/2xmqax7numnm](https://doi.org/10.63317/2xmqax7numnm) · [OpenAlex](https://openalex.org/W7160076919)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Most bias benchmarks for Chinese (e.g., CHBias, CBBQ, McBE) target text, and vision fairness datasets such as FairFace or VISOGENDER deal with single faces or occupational pictures. None of them covers the full social scenes in school textbooks. The authors built CANVAS (Chinese ANnotated Visual And Social scenes) from Chinese elementary science textbooks to fill this gap. For every person shown in an illustration, the dataset records demographic traits and social role. It also records interaction cues: who starts a conversation, who handles learning materials, who draws others' gaze, and who holds a marginal position. Labels were produced in two steps. A vision-language model first drafted structured annotations, and trained human coders then checked and corrected them, following a codebook grounded in sociocultural learning theory. The first release covers only Grade 6. A case study on gender uses it to show how scene-level and interactional labels go beyond simple head counts, for example by revealing boys more often in hands-on central roles. The resource supports fairness research spanning NLP, computer vision and education.

## Key points
- The source collection has 96 science textbooks for Grades 1–6 from eight publishers, including PEP, Jiangsu, ESE, Daxiang, Guangdong, Hebei, Hunan and Qingdao63.
- The annotated Grade 6 subset has 16 textbooks and 1,213 pages. Of these, 539 pages depict people, with 3,359 individual characters labeled.
- The schema has five modules:
  - demographics
  - the social context and the activity under way
  - the format of the interaction
  - objects in the scene and who can reach or use them
  - power dynamics, such as eye level, gaze focus, airtime, and peripheral roles
- Five coders took part in training and tryout, and three of them continued to calibration and inter-coder reliability checks.
- The analyses are meant as illustrations only. Labels for the other grades are planned for later releases.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

Fields: [Linguistics & languages](https://md.research.elamur.ai/free/en/shelves/linguistics/)

Topics: [Vision & language](https://md.research.elamur.ai/free/en/topics/multimodal/)

## Cite

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

```bibtex
@inproceedings{zhu2026canvas,
  title = {{CANVAS: A Multimodal Dataset of Chinese Textbook Images for Bias and Representation Analysis}},
  author = {Zhu, Haotian and Yu, Kefan and Li, Min},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/2xmqax7numnm},
  url = {https://doi.org/10.63317/2xmqax7numnm}
}
```

- [← previous CANDLE: CTC-based Arabic Noisy-character Deduplication using a Lightweight Encoder](https://md.research.elamur.ai/free/en/papers/alasmary-2026-candle-ctc-based-arabic-noisy/)
- [next → CATS: An Annotation Scheme of Causality and Temporal Structure](https://md.research.elamur.ai/free/en/papers/yu-2026-cats-annotation-scheme-causality-temporal/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/4/)

## References

19 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Anjali Adukia, Alex Eble, Emileigh Harrison, Hak-izumwami Birali Runesha, and Teodora Szasz. 2023. What we teach about race and gender: Representation in images and text of children books*. The Quarterly Journal of Economics, 138(4):2225–2285. [find](https://scholar.google.com/scholar?q=Anjali+Adukia%2C+Alex+Eble%2C+Emileigh+Harrison%2C+Hak%2Dizumwami+Birali+Runesha%2C+and+Teodora+Szasz%2E+2023%2E+What+we+teach+about+race+and+gender%3A+Representation+in+images+and+text+of+children+books%2A%2E+The+Quarte)
2. Anissa M Bettayeb, Manar Abu Talib, Al Zahraa Sobhe Altayasinah, and Fatima Dakalbab. 2024. Exploring the impact of chatgpt: conversational ai in education. In Frontiers in Education, volume 9, page 1379796. Frontiers Media SA. [find](https://scholar.google.com/scholar?q=Anissa+M+Bettayeb%2C+Manar+Abu+Talib%2C+Al+Zahraa+Sobhe+Altayasinah%2C+and+Fatima+Dakalbab%2E+2024%2E+Exploring+the+impact+of+chatgpt%3A+conversational+ai+in+education%2E+In+Frontiers+in+Education%2C+volume+9%2C+page+1)
3. Tolga Bolukbasi, Kai-Wei Chang, James Y. Zou, Venkatesh Saligrama, and Adam Tauman Kalai. 2016. Man is to computer programmer as woman is to homemaker? debiasing word embeddings. In Neural Information Processing Systems. [find](https://scholar.google.com/scholar?q=Tolga+Bolukbasi%2C+Kai%2DWei+Chang%2C+James+Y%2E+Zou%2C+Venkatesh+Saligrama%2C+and+Adam+Tauman+Kalai%2E+2016%2E+Man+is+to+computer+programmer+as+woman+is+to+homemaker%3F+debiasing+word+embeddings%2E+In+Neural+Information) [our card](https://md.research.elamur.ai/free/en/papers/bolukbasi-2016-man-computer-programmer-woman-homemaker/)
4. Jerome S. Bruner. 1986. Actual minds, possible worlds. Harvard University Press, Cambridge, MA. [find](https://scholar.google.com/scholar?q=Jerome+S%2E+Bruner%2E+1986%2E+Actual+minds%2C+possible+worlds%2E+Harvard+University+Press%2C+Cambridge%2C+MA%2E)
5. Siobhan Mackenzie Hall, Fernanda Gonçalves Abrantes, Hanwen Zhu, Grace Sodunke, Aleksandar Shtedritski, and Hannah Rose Kirk. 2023. Visogender: A dataset for benchmarking gender bias in image-text pronoun resolution. [find](https://scholar.google.com/scholar?q=Siobhan+Mackenzie+Hall%2C+Fernanda+Gon%C3%A7alves+Abrantes%2C+Hanwen+Zhu%2C+Grace+Sodunke%2C+Aleksandar+Shtedritski%2C+and+Hannah+Rose+Kirk%2E+2023%2E+Visogender%3A+A+dataset+for+benchmarking+gender+bias+in+image%2Dtext+pro)
6. Yufei Huang and Deyi Xiong. 2024. CBBQ: A Chinese bias benchmark dataset curated with human-AI collaboration for large language models. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 2917–2929, Torin… [find](https://scholar.google.com/scholar?q=Yufei+Huang+and+Deyi+Xiong%2E+2024%2E+CBBQ%3A+A+Chinese+bias+benchmark+dataset+curated+with+human%2DAI+collaboration+for+large+language+models%2E+In+Proceedings+of+the+2024+Joint+International+Conference+on+Com)
7. Kazi Md Mukitul Islam and M Niaz Asadullah. 2018. Gender stereotypes and education: A comparative content analysis of malaysian, indonesian, pakistani and bangladeshi school textbooks. PloS one, 13(1):e0190807. [find](https://scholar.google.com/scholar?q=Kazi+Md+Mukitul+Islam+and+M+Niaz+Asadullah%2E+2018%2E+Gender+stereotypes+and+education%3A+A+comparative+content+analysis+of+malaysian%2C+indonesian%2C+pakistani+and+bangladeshi+school+textbooks%2E+PloS+one%2C+13%281%29)
8. Kimmo Kärkkäinen and Jungseock Joo. 2021. Fairface: Face attribute dataset for balanced race, gender, and age for bias measurement and mitigation. In 2021 IEEE Winter Conference on Applications of Computer Vision (WACV), pages 1547–1557. [find](https://scholar.google.com/scholar?q=Kimmo+K%C3%A4rkk%C3%A4inen+and+Jungseock+Joo%2E+2021%2E+Fairface%3A+Face+attribute+dataset+for+balanced+race%2C+gender%2C+and+age+for+bias+measurement+and+mitigation%2E+In+2021+IEEE+Winter+Conference+on+Applications+of+Com)
9. Tian Lan, Xiangdong Su, Xu Liu, Ruirui Wang, Ke Chang, Jiang Li, and Guanglai Gao. 2025. McBE: A multi-task Chinese bias evaluation benchmark for large language models. In Findings of the Association for Computational Linguistics: ACL 2025, pages 6033–6056, Vienna, Austria. Association for Computati… [find](https://scholar.google.com/scholar?q=Tian+Lan%2C+Xiangdong+Su%2C+Xu+Liu%2C+Ruirui+Wang%2C+Ke+Chang%2C+Jiang+Li%2C+and+Guanglai+Gao%2E+2025%2E+McBE%3A+A+multi%2Dtask+Chinese+bias+evaluation+benchmark+for+large+language+models%2E+In+Findings+of+the+Association+)
10. Yaron Meron and Yasemin Tekmen Araci. 2023. Artificial intelligence in design education: evaluating chatgpt as a virtual colleague for postgraduate course development. Design Science, 9:e30. [find](https://scholar.google.com/scholar?q=Yaron+Meron+and+Yasemin+Tekmen+Araci%2E+2023%2E+Artificial+intelligence+in+design+education%3A+evaluating+chatgpt+as+a+virtual+colleague+for+postgraduate+course+development%2E+Design+Science%2C+9%3Ae30%2E)
11. OpenAI. 2025. Gpt-5 api. [https://platform.openai.com/](https://platform.openai.com/). Large language model. [↗](https://platform.openai.com/)
12. Barbara Rogoff. 2003. The cultural nature of human development. Oxford University Press, Oxford. [find](https://scholar.google.com/scholar?q=Barbara+Rogoff%2E+2003%2E+The+cultural+nature+of+human+development%2E+Oxford+University+Press%2C+Oxford%2E)
13. Ruicong Sang. 2023. Analysis of gender images in elementary school mathematics textbooks of jiangsu education press. Lecture Notes in Education Psychology and Public Media, 15:88–95. [find](https://scholar.google.com/scholar?q=Ruicong+Sang%2E+2023%2E+Analysis+of+gender+images+in+elementary+school+mathematics+textbooks+of+jiangsu+education+press%2E+Lecture+Notes+in+Education+Psychology+and+Public+Media%2C+15%3A88%E2%80%9395%2E)
14. Luthfi Mahdya Susanti, Nunung Suryati, and Utari Praba Astuti. 2021. Gender Inequality and Education: A Content Analysis of Indonesian EFL Textbook. Ph.D. thesis, State University of Malang. [find](https://scholar.google.com/scholar?q=Luthfi+Mahdya+Susanti%2C+Nunung+Suryati%2C+and+Utari+Praba+Astuti%2E+2021%2E+Gender+Inequality+and+Education%3A+A+Content+Analysis+of+Indonesian+EFL+Textbook%2E+Ph%2ED%2E+thesis%2C+State+University+of+Malang%2E)
15. Teodora Szasz, Emileigh Harrison, Ping-Jung Liu, Ping-Chang Lin, Hakizumwami Birali Runesha, and Anjali Adukia. 2022. Measuring representation of race, gender, and age in children’s books: Face detection and feature classification in illustrated images. In 2022 IEEE/CVF Winter Conference on Applicat… [find](https://scholar.google.com/scholar?q=Teodora+Szasz%2C+Emileigh+Harrison%2C+Ping%2DJung+Liu%2C+Ping%2DChang+Lin%2C+Hakizumwami+Birali+Runesha%2C+and+Anjali+Adukia%2E+2022%2E+Measuring+representation+of+race%2C+gender%2C+and+age+in+children%E2%80%99s+books%3A+Face+detect)
16. Lev S. Vygotsky. 1978. Mind in society: The development of higher psychological processes. Harvard University Press, Cambridge, MA. [find](https://scholar.google.com/scholar?q=Lev+S%2E+Vygotsky%2E+1978%2E+Mind+in+society%3A+The+development+of+higher+psychological+processes%2E+Harvard+University+Press%2C+Cambridge%2C+MA%2E)
17. Na Wang, Min Li, and Duo Dong. 2019. Measuring equitable contextualized items in science assessment. In Annual Meeting of the American Educational Research Association (AERA), Toronto, Canada. [find](https://scholar.google.com/scholar?q=Na+Wang%2C+Min+Li%2C+and+Duo+Dong%2E+2019%2E+Measuring+equitable+contextualized+items+in+science+assessment%2E+In+Annual+Meeting+of+the+American+Educational+Research+Association+%28AERA%29%2C+Toronto%2C+Canada%2E)
18. Jiaxu Zhao, Meng Fang, Zijing Shi, Yitong Li, Ling Chen, and Mykola Pechenizkiy. 2023. CHBias: Bias evaluation and mitigation of Chinese conversational language models. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 13538–135… [find](https://scholar.google.com/scholar?q=Jiaxu+Zhao%2C+Meng+Fang%2C+Zijing+Shi%2C+Yitong+Li%2C+Ling+Chen%2C+and+Mykola+Pechenizkiy%2E+2023%2E+CHBias%3A+Bias+evaluation+and+mitigation+of+Chinese+conversational+language+models%2E+In+Proceedings+of+the+61st+Annu)
19. Haotian Zhu, Kexin Gao, Fei Xia, and Mari Ostendorf. 2024. Disagreeable, slovenly, honest and un-named women? investigating gender bias in English educational resources by extending existing gender bias taxonomies. In Proceedings of the 5th Workshop on Gender Bias in Natural Language Processing (GeB… [find](https://scholar.google.com/scholar?q=Haotian+Zhu%2C+Kexin+Gao%2C+Fei+Xia%2C+and+Mari+Ostendorf%2E+2024%2E+Disagreeable%2C+slovenly%2C+honest+and+un%2Dnamed+women%3F+investigating+gender+bias+in+English+educational+resources+by+extending+existing+gender+bi)