The Development of a Natural Language Generation System For Personalized e-Health Information C. DiMarco a , H. Dominic Covvey b , P. Bray c , D. Cowan a,d , V. DiCiccio a , E. Hovy e , J. Lipa c , D. Mulholland d a David R. Cheriton School of Computer Science, University of Waterloo, Waterloo, Ontario, Canada b Waterloo Institute for Health Informatics Research, University of Waterloo, Waterloo, Ontario, Canada c Division of Plastic Surgery, Faculty of Medicine, University of Toronto, Toronto, Ontario, Canada d Conputer Systems Group, University of Waterloo, Waterloo, Ontario, Canada e Information Sciences Institute, University of Southern California, USA Abstract Personalization—adaptivity to the individual—is becoming an essential component of any com- puter-based system. In e-health systems, person- alization of health information is emerging as a key factor in the trend to patient-centric care. Patient-centric healthcare aims to engage patients in their treatment to promote greater compliance and satisfaction with their therapeutic regimens, resulting in both better patient outcomes and re- duced healthcare costs. We have developed a prototype Web-based Natural Language Genera- tion system for the authoring and subsequent per- sonalization of patient education materials. Our initial domain of application is reconstructive breast surgery, but our Natural Language software tools and authoring methodologies are generally applicable to all medical interventions. Keywords: Patient Education; Personalized e-Health; Natural Language Generation; Artificial Intelligence Introduction E-health services are playing an increasingly im- portant role in health-care management by pro- viding relevant and timely information to patients about their medical care. An important factor in the rapid growth of online health services is the trend in health management to patient-centric health care: patient-centric care aims to involve the patient directly in the medical decision- making process by providing better access to the relevant information that patients need to under- stand their medical condition and to enable them to make more-informed decisions about their pre- scribed treatment. An important new approach in patient-centric care is through ‘Information Therapy’: prescribing the right information to the right person at the right time. Ideally, Information Therapy promises to deliver specific medical information to individual patients at just the right time to assist with deci- sion-making or behaviour changes [5]. So far, however, the kinds of information delivered through specific patient portals or other individu- alized ‘information prescriptions’ is still quite generic (Claudette DeLenardo, Program Director, My CARE Source Patient Portal, Grand River Hospital, personal communication). An effective means of providing patient-centric healthcare would be through the personalization of health information: with individually tailored information, the patient would be both better- educated about their specific condition and better able to make informed decisions. As a key ele- ment in e-health services, personalized health education has great potential to provide relevant, patient-specific information that would integrate the clinical process with ‘anywhere, anytime’ healthcare delivery to produce both better- informed patients and enhanced patient outcomes. Some improvements in patient outcomes we may hope to see: greater compliance and satisfaction with therapeutic regimens, fewer complications, shorter hospital stays, and fewer return visits.