Current News

IMSC researchers developing novel language translation system for medical exams

Language Translation
Media (2.22 minutes)
quicktime | winmedia

IMSC key investigator Prof. Shri Narayanan and his team are developing a novel language translation system that allows English-speaking doctors to communicate with Persian-speaking patients about their medical concerns.

"Our system uses new technology to allow real-time spoken language communication between health professionals and patients," according to Prof. Narayanan, an associate professor of electrical engineering, computer science and linguistics who heads the Speech Analysis and Interpretation Laboratory (SAIL).

"Presently, most automatic translators are restrictive in the sense that people can speak only a fixed set of phrases," he said. "But our system allows for natural two-way dialog between people speaking two different languages."

The system, called the Transonics Spoken Dialog Translator, is being developed for the Defense Advanced Research Projects Agency (DARPA).

Prof. Narayanan said the system is in its first stages of development, but has already provided a compelling proof of concept. However, he pointed out, the creation of a truly context-aware translator that can be easily and rapidly ported to other domains and languages is an ongoing research quest.

Teams at other university and industry sites are building medical translation systems for Mandarin Chinese, Pashto and Thai for the DARPA project.

In using the laptop-based system, the doctor and patient each wear a headset with microphones and earphones. The doctor asks a question in English, and the computer translates on the fly with the patient hearing the question in Persian. When the patient answers in Persian, the computer translates the response into English immediately. They hear their questions and answers via a computer-generated voice.

Prof. Narayanan pointed out that the system has been evaluated for basic usability by English-speaking doctors and nurses with Persian speakers acting as patients. "Informal results indicate that participants can sometimes communicate quite effectively using our system," he said.

The system uses a number of cutting-edge technologies, including computer speech recognition, bi-directional language translation, dialog tracking and speech synthesis, according to Dr. Panayiotis Georgiou, research faculty in electrical engineering from IMSC and an investigator on the project.

Dr. Georgiou also said the development of the system involves extensive data collection, including collection of a corpus of interactions between USC medical students and actors portraying patients with different ailments.

The Transonics team is a partnership between USC and industry collaborator HRL Laboratories. USC team members, in addition to Prof. Narayanan and Dr. Georgiou, include Dr. Kevin Knight and Dr. Daniel Marcu, research faculty in computer science from USC's Information Sciences Institute, and Dr. David Traum, research faculty in computer science from USC's Institute for Creative Technologies.

Key HRL investigators include USC alumni Dr. Robert Belvin and Howard Neely, and Cheryl Hein. Ph.D. students in electrical engineering and computer science who have made key contributions include Emil Ettellaie, Dagen Wang, Sudeep Ghande, Ananthakrishnan Shankar and Murtaza Bulut.

View the video showing a doctor-patient interview in Quicktime or Windows Media formats. For more information on the system, visit