Elamur Research⌕
☰
home / papers / 2026 / Addressing Domain Shift in Health Coaching Note Analysis through Factorized Synthetic Data Generation

Addressing Domain Shift in Health Coaching Note Analysis through Factorized Synthetic Data Generation

Paper · 2026 · Michael Tänzer, Iva Bojic, Ashwini Yuvraj Lawate et al.

Summary

Extracting SMART behavioral goals from free-text health coaching notes could make it possible to monitor coaching programs at scale. Annotated data for this task is scarce, however, and notes differ considerably from one program to another. The authors built a new English corpus of 157 notes from a Singapore coaching program and annotated it for goal spans. Models trained only on SMARTSpan, the single public corpus, lose up to 30 F1 points on this new data, and t-SNE plots show that the two corpora occupy separate regions. To close this gap, the authors propose a factorized generation pipeline. It treats note variation as three largely separate factors: coach documentation structure, goal content, and patient persona. Gemini 3.0 Pro is used to estimate priors for each factor from a small in-domain seed set. New combinations are then sampled, an LLM writes labeled synthetic notes, and cycle-consistency filtering checks those labels. With only 57 in-domain training notes, this augmentation beats rephrasing and back-translation on both exact-match and partial-match F1. This suggests that augmentation should model the target domain's structure rather than just its wording.

Key points

  • New corpus: 157 notes from 56 clients and 3 coaches, containing 331 goals. Each note is labeled by two annotators, with word-level Cohen's κ = 0.94 and goal-level F1 = 0.79. Notes are about three times longer than SMARTSpan notes (469 vs 161 words on average).
  • Domain shift: SpanQualifier-based models trained on SMARTSpan lose substantial exact-match F1 on the new data. BERT-large-cased falls from about 90 to 62, and DeBERTa-v3-large from about 98 to 72.
  • Structure modeling: Coach structure is represented as a Markov chain over nine section types, smoothed with Kneser-Ney (d = 0.75). Goal attributes follow the Hessler et al. taxonomy of five goal types.
  • Ablation findings: Augmentation helps only when it targets the in-domain distribution. The pipeline also improves results when applied within SMARTSpan itself.
  • Human evaluation: Raters found the synthetic notes structurally faithful. When they could spot synthetic notes, it was because of surface artifacts, not flaws in content or organization. Code and generated data are to be released on GitHub (cl4health-factorized-augmentation).

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

🎓Nanyang Technological UniversitySingapore · Singapore156 works · 25,644 citations🎓Imperial College LondonLondon · United Kingdom48 works · 12,181 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

Cite

Cite as Michael Tänzer, Iva Bojic, Ashwini Yuvraj Lawate, Andy Hau Yan Ho and Andy Khong. 2026. Addressing Domain Shift in Health Coaching Note Analysis through Factorized Synthetic Data Generation. Proceedings of the Third Workshop on Patient-Oriented Language Processing (CL4Health) @ LREC 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/3a5koirj2a6s

@inproceedings{tanzer2026addressing,
  title = {{Addressing Domain Shift in Health Coaching Note Analysis through Factorized Synthetic Data Generation}},
  author = {Michael Tänzer, Iva Bojic, Ashwini Yuvraj Lawate, Andy Hau Yan Ho and Andy Khong},
  booktitle = {Proceedings of the Third Workshop on Patient-Oriented Language Processing (CL4Health) @ LREC 2026},
  year = {2026},
  doi = {10.63317/3a5koirj2a6s},
  url = {https://doi.org/10.63317/3a5koirj2a6s}
}

References

25 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Zeynab Bahrami, Atena Heidari, and Jacquelyn Cranney. 2022. Applying smart goal intervention leads to greater goal attainment, need satisfaction and positive affect. International Journal of Mental Health Promotion, 24(6). find
  2. Zixian Huang, Jiaying Zhou, Chenxu Niu, and Gong Cheng. 2023. Spans, not tokens: A span-centric model for multi-span reading comprehension. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management (CIKM), pages 874–884. find
  3. Iva Bojic, Qi Chwen Ong, Stephanie Hilary Xinyi Ma, Lin Ai, Zheng Liu, Ziwei Gong, Julia Hirschberg, Andy Hau Yan Ho, and Andy WH Khong. 2025. SMARTMiner: Extracting and evaluating SMART goals from low-resource health coaching notes. In Findings of the Association for Computational Linguistics: EMNL… find
  4. Kirsi Kivelä, Satu Elo, Helvi Kyngäs, and Maria Kääriäinen. 2014. The effects of health coaching on adult patients with chronic diseases: a systematic review. Patient education and counseling, 97(2):147–157. find
  5. Julia Bowman, Lise Mogensen, Elisabeth Marsland, and Natasha Lannin. 2015. The development, content validity and inter-rater reliability of the smart-goal evaluation method: A standardised method for evaluating clinical goals. Australian occupational therapy journal, 62(6):420–427. find
  6. Reinhard Kneser and Hermann Ney. 1995. Improved backing-off for m-gram language modeling. In 1995 international conference on acoustics, speech, and signal processing, volume 1, pages 181–184. IEEE. find
  7. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 conference of the North American chapter of the ACL: human language technologies, volume 1, pages 4171–4186.
  8. Jinhyuk Lee, Feiyang Chen, Sahil Dua, Daniel Cer, Madhuri Shanbhogue, Iftekhar Naim, Gustavo Hernández Ábrego, Zhe Li, Kaifeng Chen, Henrique Schechter Vera, et al. 2025. Gemini embedding: Generalizable embeddings from gemini. arXiv preprint arXiv:2503.07891.
  9. George T Doran. 1981. There’s a smart way to write managements’s goals and objectives. Management review, 70(11). find
  10. Haonan Li, Martin Tomko, Maria Vasardani, and Timothy Baldwin. 2022. Multispanqa: A dataset for multi-span question answering. In Proceedings of the 2022 Conference of the North American Chapter of the ACL: Human Language Technologies, pages 1250–1260. find
  11. Susan A Flocke and Kurt C Stange. 2004. Direct observation and patient recall of health behavior advice. Preventive medicine, 38(3):343–349. find
  12. Itika Gupta, Barbara Di Eugenio, Brian Ziebart, Aiswarya Baiju, Bing Liu, Ben Gerber, Lisa Sharp, Nadia Nabulsi, and Mary Smart. 2020. Human-human health coaching via text messages: Corpus, annotation, and analysis. In Proceedings of the 21th Annual Meeting of the Special Interest Group on Discourse… find
  13. Ilya Loshchilov and Frank Hutter. 2017. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101. find
  14. Croía Loughnane, Justin Laiti, Róisín O’Donovan, and Pádraic J Dunne. 2025. Systematic review exploring human, ai, and hybrid health coaching in digital health interventions: trends, engagement, and lifestyle outcomes. Frontiers in Digital Health, 7:1536416. find
  15. Itika Gupta, Barbara Di Eugenio, Brian D Ziebart, Bing Liu, Ben S Gerber, and Lisa K Sharp. 2021. Summarizing behavioral change goals from sms exchanges to support health coaches. In Proceedings of the 22nd annual meeting of the special interest group on discourse and dialogue, pages 276–289. find
  16. Ganeshan Malhotra, Abdul Waheed, Aseem Srivastava, Md Shad Akhtar, and Tanmoy Chakraborty. 2022. Speaker and time-aware joint contextual learning for dialogue-act classification in counselling conversations. In Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining,…
  17. Yue Zhou, Barbara Di Eugenio, Brian Ziebart, Lisa Sharp, Bing Liu, and Nikolaos Agadakos. 2024. Modeling low-resource health coaching dialogues via neuro-symbolic goal summarization and text-units-text generation. In Proceedings of the 2024 Joint International Conference on Computational Linguistics… find
  18. National Library of Medicine. 2024. Medical subject headings (mesh) descriptor: Sociodemographic factors. Accessed: 2026-02-19. find
  19. Jeanette M Olsen and Bonnie J Nesbitt. 2010. Health coaching to improve healthy lifestyle behaviors: an integrative review. American journal of health promotion, 25(1):e1–e12. find
  20. Jun-Yan Zhu, Taesung Park, Phillip Isola, and Alexei A Efros. 2017. Unpaired image-to-image translation using cycle-consistent adversarial networks. In Proceedings of the IEEE international conference on computer vision, pages 2223– 2232. find
  21. Zhiyang Qi, Takumasa Kaneko, Keiko Takamizo, Mariko Ukiyo, and Michimasa Inaba. 2025. Kokorochat: A japanese psychological counseling dialogue dataset collected via role-playing by trained counselors. arXiv preprint arXiv:2506.01357. find
  22. Statista. 2025. Most common languages used on the internet as of january 2025, by share of websites. Accessed: 2026-02-19. find
  23. Example Health Coaching Notes Figures 4 and 5 present two real and two synthetically generated health coaching notes, respectively. Goal spans are highlighted in green. The synthetic notes were produced by the factorised generation pipeline, sampling structure, goal attributes, and persona independe… find
  24. Laurens van der Maaten and Geoffrey Hinton. 2008. Visualizing data using t-SNE. Journal of Machine Learning Research, 9(86):2579–2605. find
  25. Anne M Wallace, Matthew T Bogard, and Susan M Zbikowski. 2018. Intrapersonal variation in goal setting and achievement in health coaching: cross-sectional retrospective analysis. Journal of Medical Internet Research, 20(1):e32. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tanzer-2026-addressing-domain-shift-health-coaching/