TransPHorm: Machine Translation for Health

TransPHorm

Translation of Public Health Information for a Diverse Public

Overview

The vast majority of health information on the Internet is available only in English. Yet, for the over 46 million people living in the US with limited English proficiency (LEP), obtaining accurate and up-to-date health information can be very difficult.

The demand for translation of multi-lingual health materials is not being met by health departments in part because of the time and resources it takes to create high quality translations. Machine translation (MT) technology holds promise for creating health translations efficiently and accurately. However, the quality of automated translations for public health materials is currently poor.

To address this problem, Anne Turner, MD, MLIS, MPH, Assistant Professor, joint, Health Services and Medical Education and Biomedical Informatics and Katrin Kirchhoff, PhD, Associate Professor, Electrical Engineering have received funding from the National Library of Medicine (NLM) to investigate the use of MT technology to improve the time and costs of producing translations of health promotion materials from English to Spanish and Vietnamese. The goal is to improve automated translations of public health materials in terms of costs, turnaround time, and quality.

Methods

Over the five year period of the National Library of Medicine grant, we will:

  • Use qualitative methods to investigate current public health translation processes to better understand translation workflow and the time and costs of creating multi-lingual materials
  • Apply cognitive work analysis to describe the strengths and weaknesses of current MT technology
  • Develop novel methods of improving MT in the domain of public health
  • Systematically compare MT with manual translations in terms of time, quality, and costs

Notes from the Field

  • Interviewed 21 staff from Public Health - Seattle & King County to identify current translation activities and processes
  • Developing a survey for local health agencies in Health and Human Services Region 10 (Washington, Oregon, Idaho, and Alaska) to gather information on modes of communication/dissemination and translation of health promotion materials
  • Conducted an online study to identify the impact of different types of machine translation errors on users' preferences.

Researcher Team

Anne Turner, MD, MLIS, MPH and Katrin Kirchhoff, PhD, Principal Investigators
Megumu Brownstein, MSW, MPH, Research Coordinator
Adrian Laurenzi, BCc, Research Assistant
Daniel Capurro, MD, Fullbright Scholar, Research Assistant
Tina Neogi, MD, NRSA Fellow, Research Assistant
Hilary Karasz, PhD, PHSKC Co-Investigator
Julia Cordero, PHSKC Liaison

For questions, please contact Megumu Brownstein at megumu@uw.edu.

Learn more about why local health departments want to participate in this study.