Elamur Research⌕
☰
home / papers / 2007 / Theoretical and methodological issues regarding the use of language technologies for patients with limited English proficiency

Theoretical and methodological issues regarding the use of language technologies for patients with limited English proficiency

Paper · 2007 · Harold Somers

Summary

This position paper asks how language technology can help healthcare providers and patients who share no common language, using patients with limited English proficiency (LEP) as the case. Somers maps the whole pathway to healthcare: self-diagnosis online, booking appointments, history-taking, consultation, pharmacy, and follow-up care. He shows that each step calls for a different technology, such as SLT, controlled-language MT, text simplification or computer-based interviewing. He makes two main arguments. First, design should centre on users, and the patient may be the device's main and repeated user, not the doctor. Second, building SLT for under-resourced languages costs so much that simpler tools or deliberate "cheats" are often wiser. Evidence comes from small studies:

  • a Chinese–English pipeline, where ASR was the weak link;
  • poor OCR of prescription labels;
  • a laptop/tablet tool for Somali asthma patients built on predetermined questions with recorded answers.

The tool got very high satisfaction but could not handle off-script talk. The paper's message is that user needs, not technical ambition, should guide the choice of technology.

Key points

  • Wang (2007) chained commercial Chinese ASR, Chinese–English MT and English synthesis. About 70% of ASR+MT outputs were interpreted correctly, falling to 62% with speech output, while MT alone was 97% understandable. In Somers & Sugita (2003), by contrast, Japanese ASR was reliable and MT was the weak step.
  • Interim workarounds for under-resourced languages:
    • Somali output produced with a German speech synthesizer;
    • an English ASR system coaxed into recognising a small Urdu vocabulary, which worked very well when picking among a fixed set of responses.
  • Ghobadi (2007) scanned prescription labels with a handheld scanner and got largely unusable text, because of cheap printing and curved containers.
  • The Somali asthma tool offered 69 bilingual English/Somali questions with pre-recorded human speech and symbol-based answers. Its evaluation involved 26 Somali patients alongside six clinicians, a mix of GPs and asthma nurses.
    • All 26 simulated consultations finished, and in 20 every question was answered.
    • Clinicians were least satisfied with eliciting the patient's worries (42%), building a relationship (69%) and achieving the desired outcome (65%).
    • Patients rated the tablet at 91% and the stylus at 100%, but the laptop and mousepad at only 43%.
  • Pharmacy dosage instructions are presented as a natural fit for controlled-language MT, about as complex as weather bulletins.

Interesting ideas

  • I1. Testing each stage of a pipeline separately, then in combination, shows which component actually loses the meaning, rather than giving one blended score.
  • I2. A weak recognizer becomes reliable when it only has to choose among a small set of expected responses instead of decoding open input.
  • I3. Translation quality was judged by whether a person could identify the intended meaning of an answer to a known question, not by matching surface wording.
  • I4. For predictable, narrow content such as dosage instructions, controlled language and fixed templates are safer than full open-domain translation.
  • I5. Designing for two users with very different skills at once, and adapting the interface to each user's profile, makes a tool usable in practice.

For Mova

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

  • I1 (learn): Mova's evaluation could score tokens, tree, frames and world separately on the same items, and report where correct answers first break, to decide which stage to repair.
  • I2 (answer): for school tests and multiple-choice QA, the reader could rank candidate answers against the world of the text rather than generating free answers, which keeps the choice explainable.
  • I3 (back): the text → graph → text compressor could be evaluated by whether regenerated text still answers the same probe questions, instead of by string overlap.
  • I4 (back): generation could switch to template-based realisation for domains with fixed patterns (measures, quantities, instructions), with free generation reserved for open text.
  • I5 (frames): dialogue frames could record each speaker's role and assumed knowledge separately, so speech acts and implicatures are interpreted per participant rather than for one generic speaker.

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, topics and concepts

📚 Shelves 6The 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 worksHealth Professions · 69 worksGeneral Health Professions · 14 works

🧭 Topics 2Research subjects and languages the work is about.Machine translation · 1,087 worksEnglish · 2,263 works

Cite

Cite as Harold Somers. 2007. Theoretical and methodological issues regarding the use of language technologies for patients with limited English proficiency. Proceedings of the 11th Conference on Theoretical and Methodological Issues in Machine Translation of Natural Languages: Papers.

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

@inproceedings{somers2007theoretical,
  title = {{Theoretical and methodological issues regarding the use of language technologies for patients with limited English proficiency}},
  author = {Harold Somers},
  booktitle = {Proceedings of the 11th Conference on Theoretical and Methodological Issues in Machine Translation of Natural Languages: Papers},
  year = {2007},
  url = {https://aclanthology.org/2007.tmi-papers.25/}
}

References

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

  1. Bachman JW (2003) The patient-computer interview: a neglected tool that can aid the clinician. Mayo Clinic Proceedings 78, 67–78. ↗
  2. Barclay L (2007) Pharmacies may not always translate prescription labels for non-English speaking patients. [Report on presentation by L Weiss at Society for General Internal Medicine 2007 Annual Meeting, Toronto, Ont. Abstract 172022]. Medscape Medical News, April 27, 2007. Available at: http://www.medscape.com/viewarticle/555840. Accessed 14 May 2007. ↗
  3. Besacier L, Le V-B, Boitet C, Berment V (2006) ASR and translation for under-resourced languages. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Toulouse, France, pp. V-1221–4. find
  4. Bischoff A, Bovier PA, Isah R, Francoise G, Ariel E, Louis L (2003) Language barriers between nurses and asylum seekers: their impact on symptom reporting and referral. Social Science in Medicine 57, 503–12. ↗
  5. Black AW, Brown RD, Frederking RF, Singh R, Moody J, Steinbrecher E (2002) TONGUES: Rapid development of a speech-to-speech translation system. Proceedings of HLT 2002: Second International Conference on Human Language Technology Research, San Diego, California, pp. 183–6. find
  6. Bouillon P, Flores G, Starlander M, Chatzichrisafis N, Santaholma M, Tsourakis N, Rayner M, Hockey BA (2007) A bidirectional grammar-based medical speech translator. Proceedings of the Workshop on Grammar-based approaches to spoken language processing, Prague, Czech Republic, pp. 41–48. find
  7. Bouillon P, Rayner M, Chatzichrisafis N, Hockey BA, Santaholma M, Starlander M, Nakao Y, Kanzaki K, IsaharaH (2005) A generic multi-lingual open source platform for limited-domain medical speech translation. Proceedings of the Tenth Conference on European Association of Machine Translation, “Practical applications of machine translation”, Budapest, Hungary, pp. 50–8. find
  8. Ehsani F, Kimzey J, Master D, Park H, Sudre K (2006) Rapid development of a speech translation system for Korean. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Toulouse, France, pp. V-1225–8. find
  9. Ettelaie E, Gandhe S, Georgiou P, Belvin R, Knight K, Marcu D, Narayanan S, Traum D (2005) Transonics: A practical speech-to-speech translator for English-Farsi medical dialogues. 43rd Annual Meeting of the Association for Computational Linguistics: ACL-05 Interactive Poster and Demonstration Sessions, Ann Arbor, MI, pp. 89–92. find
  10. Evans G, Polyzoaki K, Blenkhorn P (2002) An approach to producing new languages for talking applications for use by blind people. In K Miesenberger, J Klaus, W Zagler (eds) 8th ICCHP, Computers Helping People with Special Needs, (LNCS 2398), Berlin: Springer Verlag, pp. 575–82. find
  11. Fassil Y (2002) Looking after the health of refugees. British Medical Journal 321, 59. find
  12. Flores G, Abreu M, Tomany-Korman SC (2005) Limited English proficiency, primary language at home, and disparities in children’s health care: How language barriers are measured matters. Public Health Report 120, 418–30. ↗
  13. Ghobadi B (2007) Farsi prescription labels translator. Dissertation, School of Informatics, University of Manchester. find
  14. Jacobs EA, Lauderdale DS, Meltzer D, Shorey JM, Levison W, Thisted R (2001) Impact of interpreter services on delivery of health care to limited-English-proficient patients. Journal of General Internal Medicine 16, 468–74. ↗
  15. Johnson MJ (2007) An exploration into support for communication between health care practitioners and Somalis using assistive language technology in the context of asthma consultations. PhD thesis, School of Nursing, Midwifery and Social Work/School of Informatics, University of Manchester. find
  16. Jones D, Gill P (1998) Refugees and primary care: tackling the inequalities. British Medical Journal 317, 1444–6. find
  17. Kathol A, Precoda K, Vergyri D, Wang W, Riehemann S (2005) Speech translation for low-resource languages: The case of Pashto. Interspeech’2005 – Eurospeech, Lisbon, pp. 2273–6. find
  18. Narayanan S., Ananthakrishnan S, Belvin R, Ettelaie E, Gandhe S, Ganjavi S, Georgiou PG, Hein CM, Kadambe S, Knight K, Marcu D, Neely HE, Srinivasamurthy N, Traum, D, Wang D (2004) The Transonics spoken dialogue translator: An aid for English-Persian doctor-patient interviews. In T Bickmore (ed.) Dialogue Systems for Health Communication: Papers from the 2004 Fall Symposium, Menlo Park, California: American Association for Artificial Intelligence, pp. 97–103. find
  19. Narayanan SS, Georgiou PG, Sethy A, Wang D, Bulut M, Sundaram S, Ettalaie E, Ananthakrishnan S, Franco H, Precoda K, Vergyri D, Zheng J, Wang W, Gadde RR, Graciarena M, Abrash V, Frandsen M, Richey C (2006) Speech recognition engineering issues in speech to speech translation system design for low resource languages and domains. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Toulouse, France, pp. V-1209–12. find
  20. OCR (Office of Civil Rights) (1998) Guidance memorandum: Title VI prohibition against national origin discrimination—persons with limited-English proficiency. January 29, 1998. Available at: http://www.hhs.gov/ocr/lepfinal.htm. Accessed 14 May 2007. ↗
  21. Orlovsky C (2005) Talking pill bottles let medications speak for themselves’. NurseZone.com (online magazine). Available at: www.nursezone.com/Job/DevicesandTechnology.asp?articleID=14396. Accessed 14 May 2007. find
  22. Osman L, Abdalla M, Beattie J, Ross S, Russell I, Friend J, Legge J, Douglas J (1994) Reducing hospital admissions through computer supported education for asthma patients. British Medical Journal 308, 568–71. find
  23. Precoda K, Franco H, Dost A, Frandsen M, Fry J, Kathol A, Richey C, Riehemann S, Vergyri D, Zheng J, Culy C (2004) Limited-domain speech-to-speech translation between English and Pashto. HLT-NAACL 2004, Human Language Technology Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations, Boston, MA, pp. 9–12. find
  24. Schultz T, Alexander D, Black AW, Peterson K, Suebvisai S, Waibel A (2004) A Thai speech translation system for medical dialogs. HLT-NAACL 2004, Human Language Technology Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations, Boston, MA, pp. 34–5. find
  25. Schultz T, Black AW (2006) Challenges with rapid adaptation of speech translation systems to new language pairs. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Toulouse, France, pp. V-1213–16. find
  26. Sharif I, Lo S, Ozuah PO (2006) Availability of Spanish prescription labels. Journal of Health Care for the Poor and Underserved 17, 65–9. ↗
  27. Somers H (2006) Language Engineering and the pathway to healthcare: A user-oriented view. HLT/NAACL-06 Medical Speech Translation, Proceedings of the Workshop, New York, pp. 32–9. find
  28. Somers H, Evans G, Mohamed Z (2006) Developing speech synthesis for under-resourced languages by “faking it”: An experiment with Somali. 5th International Conference on Language Resources and Evaluation, Genoa, Italy, pp. 2578–81. find
  29. Somers H, Lovel H (2003) Computer-based support for patients with limited English. Association for Computational Linguistics EACL 2003, 10th Conference of the European Chapter, Proceedings of the 7th International EAMT Workshop on MT and other language technology tools, Budapest, pp. 41–9. find
  30. Somers HL, Lovel HJ (2007) Can AAC technology facilitate communication for patients with limited English? ESRC Project Final Report, School of Informatics, University of Manchester. Available at http://www.informatics.manchester.ac.uk/~harold/ESRCfinal.pdf. Accessed 8 August 2007. ↗
  31. Somers H, Sugita Y (2003) Evaluating commercial spoken language translation software. MT Summit IX: Proceedings of the Ninth Machine Translation Summit, New Orleans, pp. 370–7. find
  32. Stewart M, Brown JB, Weston WW, McWhinney IR, McWilliam CL, Freeman TR (2003) Patient-centered medicine: Transforming the clinical method (2nd ed.). Abingdon, Oxon: Radcliffe. find
  33. Wang B (2007) Chinese to English speech translation system built from standard components. Dissertation, School of Informatics, University of Manchester. find
  34. Westberg SM, Sorensen TD (2005) Pharmacy-related health disparities experienced by non-English-speaking patients: Impact of pharmaceutical care. Journal of the American Pharmaceutical Association 45, 48–54. ↗
  35. Zhou B, Déchelotte D, Gao Y (2004) Two-way speech-to-speech translation on handheld devices. INTERSPEECH 2004 – ICSLP, 8th International Conference on Spoken Language Processing, Jeju Island, Korea, pp. 1637–40. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/somers-2007-theoretical-methodological-issues-regarding-use/