Project description

At the Hôpitaux Universitaires de Genève (HUG), Geneva's largest hospital, 52% of all patients are foreign nationals and more than 10% speak no French at all. In the context of the ongoing European refugee crisis, the medical professionals at HUG, particularly in the emergency and immigrant health service departments, often find they have no language in common with a patient. Particularly important languages are Tigrinya, Arabic and Farsi; as of September 2015, Eritreans, Syrians and Afghans make up about 60% of all new asylum seekers. Language barriers of this kind pose serious problems regarding the quality, security and equitability of health care, a phenomenon which has been the subject of detailed investigation by several teams over the last twenty years (Flores et al. "Errors in medical interpretation and their potential clinical consequences in pediatric encounters." Pediatrics 2003; Wasserman et al. "Identifying and Preventing Medical Errors in Patients With Limited English Proficiency: Key Findings and Tools for the Field". Journal for Healthcare Quality 2014 ).

The BabelDr speech translation application is designed to address these problems. The system has been specifically designed to assist in triaging of non-French-speaking patients visiting HUG's A&E department, and allows a medical professional to perform a preliminary medical examination dialogue, using a decision-tree method, to determine the nature of the patient's problem and the appropriate action to take. In the current version of the system, the patient is expected to respond non-verbally, e.g. by nodding, shaking their head, or pointing. BabelDr differs from general speech translation systems like Google Translate in several important respects. Speech recognition is performed using linguistic methods and is specialized for medical translation dialogues, which allows excellent recognition quality for phrases covered by the system without requiring speaker adaptation. The range of language covered by the system has been defined by doctors at HUG. All questions are mapped into a set of standard types that are translated in advance by expert translators at the University of Geneva, guaranteeing reliable output.

The system has been developed by the Department of Multilingual Information Processing (TIM) at the University of Geneva and HUG. The BabelDr project is financed by the Fondation privée des Hôpitaux universitaires de Genève and by Unitec.