---
title: "Structured Entity Extraction from Hawaiian — Lynch 2026"
description: "Hawaiian public television archives hold decades of on-screen name captions (chyrons) that could supply catalog metadata."
authors: ["Kelley Lynch", "Owen King", "Kyeongmin Rim", "Gabrielle Keen", "Yangyang Chen", "James Pustejovsky"]
year: 2026
venue: "Proceedings of the SIGUL 2026 Joint Workshop with ELE, EURALI, and DCLRL: Towards Inclusivity and Equality: Language Resources and Technologies for Under-Resourced and Endangered Languages"
source: "https://aclanthology.org/2026.sigul-1.17/"
doi: "10.63317/4vjma77xc4kh"
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/papers/lynch-2026-structured-entity-extraction-hawaiian-television/"
---
[Elamur Research](https://md.research.elamur.ai/) · [this page in HTML](https://research.elamur.ai/free/papers/lynch-2026-structured-entity-extraction-hawaiian-television/)

[home](https://md.research.elamur.ai/) / [papers](https://md.research.elamur.ai/free/papers/) / [2026](https://md.research.elamur.ai/free/papers/2026/page/4/) / Structured Entity Extraction from Hawaiian Television Chyrons Using Vision-Language Models

# Structured Entity Extraction from Hawaiian Television Chyrons Using Vision-Language Models

**Authors:** Kelley Lynch, Owen King, [Kyeongmin Rim](https://md.research.elamur.ai/free/authors/kyeongmin-rim/), Gabrielle Keen, Yangyang Chen, [James Pustejovsky](https://md.research.elamur.ai/free/authors/james-pustejovsky/) · **Year:** 2026 · **Venue:** Proceedings of the SIGUL 2026 Joint Workshop with ELE, EURALI, and DCLRL: Towards Inclusivity and Equality: Language Resources and Technologies for Under-Resourced and Endangered Languages · **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.sigul-1.17/) · [PDF](https://aclanthology.org/2026.sigul-1.17.pdf) · [DOI 10.63317/4vjma77xc4kh](https://doi.org/10.63317/4vjma77xc4kh) · [OpenAlex](https://openalex.org/W7166503252)

**Citations:** 0 citations (openalex-snapshot · 2026-10-14) · 1 in our library, of the 140673 papers whose bibliographies we read

## Summary
Hawaiian public television archives hold decades of on-screen name captions (chyrons) that could supply catalog metadata. It was unclear how well current vision-language models handle them, especially names from ‘Ōlelo Hawai‘i, an endangered language with almost no computational resources. The authors build HiChy, 3,925 hand-annotated chyron frames: 1,825 from PBS Hawai‘i programs (1975–2008) and 2,100 from comparable mainland U.S. public broadcasts. Each image has a transcription, the name as displayed, a "Lastname, Firstname" normalization, a list of attributes such as roles and affiliations, and a label for the name's linguistic origin. Five VLMs are tested on two tasks, plain OCR and key information extraction (KIE) into JSON, using few-shot and zero-shot prompts. Hawaiian material proves markedly harder: transcription error roughly doubles compared with mainland data. Names of Hawaiian origin remain difficult even when the comparison is restricted to a single source. Strong OCR does not guarantee good structured extraction. The study sets a baseline for AI-assisted archiving of media in under-represented languages.

## Key points
- **Dataset and labels:** HiChy covers 158 Hawaiian and 131 comparison programs. Name origin labels are Hawaiian (225), Japanese (200), both (34) and neither (3,466). They were assigned with AI assistance and verified by humans.
- **Models tested:** Qwen2.5-VL-7B, Qwen2.5-VL-32B, Gemini 3 Flash (API), Nanonets-OCR2 (3B) and SmolVLM2 (2.2B). The local models ran on one RTX A6000 (48GB), mostly as quantized GGUF via Ollama, to reflect typical archive hardware.
- **OCR gap:** For the best model, Qwen2.5-VL-7B, character error rate rises from 0.064 on mainland chyrons to 0.130 on Hawaiian ones.
- **OCR vs. KIE:** Gemini 3 Flash is the top OCR model but trails locally deployed models on KIE. The small SmolVLM2 still produces usable structured output.
- **Prompting:** Few-shot examples mainly help the model follow the name-normalization convention. The KIE metrics follow document-AI practice, e.g., ANLS/ANLS*.

*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/shelves/linguistics/)

## Cite

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

```bibtex
@inproceedings{lynch2026structured,
  title = {{Structured Entity Extraction from Hawaiian Television Chyrons Using Vision-Language Models}},
  author = {Lynch, Kelley and King, Owen and Rim, Kyeongmin and Keen, Gabrielle and Chen, Yangyang and Pustejovsky, James},
  booktitle = {Proceedings of the SIGUL 2026 Joint Workshop with ELE, EURALI, and DCLRL: Towards Inclusivity and Equality: Language Resources and Technologies for Under-Resourced and Endangered Languages},
  year = {2026},
  doi = {10.63317/4vjma77xc4kh},
  url = {https://doi.org/10.63317/4vjma77xc4kh}
}
```

- [← previous Stands to Reason: Investigating the Effect of Reasoning on Idiomaticity Detection](https://md.research.elamur.ai/free/papers/phelps-2026-stands-reason-investigating-effect-reasoning/)
- [next → Structured Legal Document Generation in India: A Model-Agnostic Wrapper Approach with VidhikDastaavej](https://md.research.elamur.ai/free/papers/nigam-2026-structured-legal-document-generation-india/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/papers/2026/page/4/)

## References

19 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. languages. In Proceedings of the 4th Workshop on Natural Language Processing for Indigenous Languages of the Americas (AmericasNLP 2024), pages 88–102, Mexico City, Mexico. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=languages%2E+In+Proceedings+of+the+4th+Workshop+on+Natural+Language+Processing+for+Indigenous+Languages+of+the+Americas+%28AmericasNLP+2024%29%2C+pages+88%E2%80%93102%2C+Mexico+City%2C+Mexico%2E+Association+for+Computation)
2. Siyu An, Ye Liu, Haoyuan Peng, and Di Yin. 2024. VKIE: The Application of Key Information Extraction on Video Text. [find](https://scholar.google.com/scholar?q=Siyu+An%2C+Ye+Liu%2C+Haoyuan+Peng%2C+and+Di+Yin%2E+2024%2E+VKIE%3A+The+Application+of+Key+Information+Extraction+on+Video+Text%2E)
3. Ali Furkan Biten, Rubèn Tito, Andrés Mafla, Lluís Gomez, Marçal Rusiñol, C.V. Jawahar, Ernest Valveny, and Dimosthenis Karatzas. 2019. Scene Text Visual Question Answering. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). [find](https://scholar.google.com/scholar?q=Ali+Furkan+Biten%2C+Rub%C3%A8n+Tito%2C+Andr%C3%A9s+Mafla%2C+Llu%C3%ADs+Gomez%2C+Mar%C3%A7al+Rusi%C3%B1ol%2C+C%2EV%2E+Jawahar%2C+Ernest+Valveny%2C+and+Dimosthenis+Karatzas%2E+2019%2E+Scene+Text+Visual+Question+Answering%2E+In+Proceedings+of+the+IEEE%2F)
4. David M. Eberhard, Gary F. Simons, and Charles D. Fennig. 2025. Ethnologue: Languages of the World, 28 edition. SIL International, Dallas. [find](https://scholar.google.com/scholar?q=David+M%2E+Eberhard%2C+Gary+F%2E+Simons%2C+and+Charles+D%2E+Fennig%2E+2025%2E+Ethnologue%3A+Languages+of+the+World%2C+28+edition%2E+SIL+International%2C+Dallas%2E)
5. Ling Fu et al. 2025. OCRBench v2: An Improved Benchmark for Evaluating Large Multimodal Models on Visual Text Localization and Reasoning. [find](https://scholar.google.com/scholar?q=Ling+Fu+et+al%2E+2025%2E+OCRBench+v2%3A+An+Improved+Benchmark+for+Evaluating+Large+Multimodal+Models+on+Visual+Text+Localization+and+Reasoning%2E)
6. Google DeepMind. 2025. Introducing Gemini 3: Our most capable models yet. [find](https://scholar.google.com/scholar?q=Google+DeepMind%2E+2025%2E+Introducing+Gemini+3%3A+Our+most+capable+models+yet%2E)
7. Zheng Huang, Kai Chen, Jianhua He, Xiang Bai, Dimosthenis Karatzas, Shijian Lu, and C.V. Jawahar. 2019. ICDAR2019 Competition on Scanned Receipt OCR and Information Extraction. In Proceedings of the International Conference on Document Analysis and Recognition (ICDAR). [find](https://scholar.google.com/scholar?q=Zheng+Huang%2C+Kai+Chen%2C+Jianhua+He%2C+Xiang+Bai%2C+Dimosthenis+Karatzas%2C+Shijian+Lu%2C+and+C%2EV%2E+Jawahar%2E+2019%2E+ICDAR2019+Competition+on+Scanned+Receipt+OCR+and+Information+Extraction%2E+In+Proceedings+of+the+I)
8. Soumya Jahagirdar, Minesh Mathew, Dimosthenis Karatzas, and C. V. Jawahar. 2023. Watching the News: Towards VideoQA Models that can Read. [find](https://scholar.google.com/scholar?q=Soumya+Jahagirdar%2C+Minesh+Mathew%2C+Dimosthenis+Karatzas%2C+and+C%2E+V%2E+Jawahar%2E+2023%2E+Watching+the+News%3A+Towards+VideoQA+Models+that+can+Read%2E)
9. Guillaume Jaume, Hazim Kemal Ekenel, and Jean-Philippe Thiran. 2019. FUNSD: A Dataset for Form Understanding in Noisy Scanned Documents. In Proceedings of the International Conference on Document Analysis and Recognition Workshops (ICDARW). [find](https://scholar.google.com/scholar?q=Guillaume+Jaume%2C+Hazim+Kemal+Ekenel%2C+and+Jean%2DPhilippe+Thiran%2E+2019%2E+FUNSD%3A+A+Dataset+for+Form+Understanding+in+Noisy+Scanned+Documents%2E+In+Proceedings+of+the+International+Conference+on+Document+Anal)
10. Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, and Saining Xie. 2022. A ConvNet for the 2020s. [find](https://scholar.google.com/scholar?q=Zhuang+Liu%2C+Hanzi+Mao%2C+Chao%2DYuan+Wu%2C+Christoph+Feichtenhofer%2C+Trevor+Darrell%2C+and+Saining+Xie%2E+2022%2E+A+ConvNet+for+the+2020s%2E)
11. Souvik Mandal, Ashish Talewar, Paras Ahuja, and Prathamesh Juvatkar. 2025. Nanonets-OCR-s. Fine-tuned from Qwen2.5-VL-3B-Instruct. [find](https://scholar.google.com/scholar?q=Souvik+Mandal%2C+Ashish+Talewar%2C+Paras+Ahuja%2C+and+Prathamesh+Juvatkar%2E+2025%2E+Nanonets%2DOCR%2Ds%2E+Fine%2Dtuned+from+Qwen2%2E5%2DVL%2D3B%2DInstruct%2E)
12. Andrés Marafioti et al. 2025. SmolVLM: Redefining small and efficient multimodal models. [find](https://scholar.google.com/scholar?q=Andr%C3%A9s+Marafioti+et+al%2E+2025%2E+SmolVLM%3A+Redefining+small+and+efficient+multimodal+models%2E)
13. Stephen Mayhew, Terra Blevins, Shuheng Liu, et al. 2024. Universal NER: A gold-standard multilingual named entity recognition benchmark. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics, pages 4322–4337. Association for Computationa… [find](https://scholar.google.com/scholar?q=Stephen+Mayhew%2C+Terra+Blevins%2C+Shuheng+Liu%2C+et+al%2E+2024%2E+Universal+NER%3A+A+gold%2Dstandard+multilingual+named+entity+recognition+benchmark%2E+In+Proceedings+of+the+2024+Conference+of+the+North+American+Cha)
14. Seunghyun Park, Seung Shin, Bado Lee, Junyeop Lee, Jaeheung Surh, Minjoon Seo, and Hwalsuk Lee. 2019. CORD: A Consolidated Receipt Dataset for Post-OCR Parsing. In Workshop on Document Intelligence at NeurIPS. [find](https://scholar.google.com/scholar?q=Seunghyun+Park%2C+Seung+Shin%2C+Bado+Lee%2C+Junyeop+Lee%2C+Jaeheung+Surh%2C+Minjoon+Seo%2C+and+Hwalsuk+Lee%2E+2019%2E+CORD%3A+A+Consolidated+Receipt+Dataset+for+Post%2DOCR+Parsing%2E+In+Workshop+on+Document+Intelligence+at)
15. David Peer, Philemon Schöpf, Volckmar Nebendahl, Alexander Rietzler, and Sebastian Stabinger. 2024. ANLS* – A Universal Document Processing Metric for Generative Large Language Models. [find](https://scholar.google.com/scholar?q=David+Peer%2C+Philemon+Sch%C3%B6pf%2C+Volckmar+Nebendahl%2C+Alexander+Rietzler%2C+and+Sebastian+Stabinger%2E+2024%2E+ANLS%2A+%E2%80%93+A+Universal+Document+Processing+Metric+for+Generative+Large+Language+Models%2E)
16. Qwen Team. 2025. Qwen2.5-VL: To See the World with Wisdom. [find](https://scholar.google.com/scholar?q=Qwen+Team%2E+2025%2E+Qwen2%2E5%2DVL%3A+To+See+the+World+with+Wisdom%2E)
17. Kyeongmin Rim, Owen C. King, Kelley Lynch, Marc Verhagen, and James Pustejovsky. 2025. A Platform for AI-Assisted Archival Metadata Generation. In Culture and Computing, pages 183–203. Springer Nature Switzerland. [find](https://scholar.google.com/scholar?q=Kyeongmin+Rim%2C+Owen+C%2E+King%2C+Kelley+Lynch%2C+Marc+Verhagen%2C+and+James+Pustejovsky%2E+2025%2E+A+Platform+for+AI%2DAssisted+Archival+Metadata+Generation%2E+In+Culture+and+Computing%2C+pages+183%E2%80%93203%2E+Springer+Nature)
18. Brendan Shillingford and Oiwi Parker Jones. 2018. Recovering missing characters in old Hawaiian writing. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 4929–4934, Brussels, Belgium. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Brendan+Shillingford+and+Oiwi+Parker+Jones%2E+2018%2E+Recovering+missing+characters+in+old+Hawaiian+writing%2E+In+Proceedings+of+the+2018+Conference+on+Empirical+Methods+in+Natural+Language+Processing%2C+page)
19. William H. Wilson and Kauanoe Kamanā. 2001. Mai loko mai o ka ’i’ini: Proceeding from a dream: The ’Aha Pūnana Leo connection in Hawaiian language revitalization. In Leanne Hinton and Ken Hale, editors, The Green Book of Language Revitalization in Practice, pages 147–176. Academic Press, San Diego… [find](https://scholar.google.com/scholar?q=William+H%2E+Wilson+and+Kauanoe+Kamana%CC%84%2E+2001%2E+Mai+loko+mai+o+ka+%E2%80%99i%E2%80%99ini%3A+Proceeding+from+a+dream%3A+The+%E2%80%99Aha+Pu%CC%84nana+Leo+connection+in+Hawaiian+language+revitalization%2E+In+Leanne+Hinton+and+Ken+Hale%2C+edit)