Medical question-answering at the point of care
Answers to clinical questions, such as questions posed by clinicians in patient treatment, often require multiple pieces of information. For example:
Q: In a patient with a generalized anxiety disorder, does cognitive behavior or relaxation therapy decrease symptoms?
Superficially, this is just a yes / no question. But the clinical outcomes of the therapies could be complicated. They could have different effects for different patient groups; some clinical trials may show they are beneficial while others don't. Answers to these questions can be obtained in the book Clinical Evidence (CE), a publication that reviews and consolidates experimental results for clinical problems; it is updated every six months. This project uses an online version of CE as the basis for answers. Thus, natural language processing is involved in two places: in the natural language queries to the system, and in searching the text of CE for answers. Simple keyword searching is quite inadequate to the task, which we frame as a problem of non-factoid question-answering (NFQA).
In her dissertation, Yun Niu proposed an approach to the problem that uses semantic class analysis as the organizing principle to answer non-factoid questions. This approach contains four major components:
1. Detecting semantic classes in questions and answer sources;
2. Identifying properties of semantic classes;
3. Question-answer matching: exploring properties of semantic classes to find relevant pieces of information;
4. Constructing answers by merging or synthesizing relevant information using relations between semantic classes.
Niu investigated NFQA in the context of clinical question answering, and focused on three semantic classes that correspond to roles in the commonly accepted PICO format of describing clinical scenarios. The three classes are: the problem of the patient, the intervention used to treat the problem, and the clinical outcome. She uses rule-based approaches to identify clinical outcomes and relations between instances of interventions in sentences.
Niu identified an important property of semantic classes -- their cores. She showed how cores of interventions, problems, and outcomes in a sentence can be extracted automatically by developing an approach exploring semi-supervised learning techniques. Another property that she analyze is polarity, an inherent property of clinical outcomes. She developed a method using a supervised learning model to automatically detect polarity of clinical outcomes.
Niu has shown that text summarization is closely related to answer construction in NFQA, and that summarization techniques can be adapted to identifying components of answers. She built a summarization system that explores a supervised classification model to extract important sentences for answer construction, and investigated the role of clinical outcome and their polarity in this task.
This research was a component of the University of Toronto EPoCare project at the .
Niu, Yun and Hirst, Graeme. “Analyzing the text of clinical literature for question answering.” In: Prince, Violaine and Roche, Mathieu (editors), Information Retrieval in Biomedicine, IGI Global, 2009, 190-220. 
Niu, Yun and Hirst, Graeme. “Identifying cores of semantic classes in unstructured text with a semi-supervised learning approach.” Proceedings, International Conference on Recent Advances in Natural Language Processing, September 2007, Borovets, Bulgaria, 418–424. 
Niu, Yun. Analysis of Semantic Classes: Toward non-factoid question answering. Ph.D. Thesis. Department of Computer Science, University of Toronto. March 2007. 
Niu, Yun; Zhu, Xiaodan; and Hirst, Graeme. “Using outcome polarity in sentence extraction for medical question-answering.” Proceedings of the American Medical Informatics Association 2006 Annual Symposium, Washington, D.C., November 2006, 599-603. 
Niu, Yun; Zhu, Xiaodan; Li, Jianhua; and Hirst, Graeme. “Analysis of polarity information in medical text.” Proceedings of the American Medical Informatics Association 2005 Annual Symposium, Washington, D.C., October 2005, 570-574. 
Niu, Yun and Hirst, Graeme. “Analysis of semantic classes in medical text for question answering.” Workshop on Question Answering in Restricted Domains at the 42nd Annual Meeting of the Association for Computational Linguistics, Barcelona, July 2004, 54-61. 
Niu, Yun; Hirst, Graeme; McArthur, Gregory; and Rodriguez-Gianolli, Patricia. “Answering clinical questions with role identification.” Proceedings, Workshop on Natural Language Processing in Biomedicine, 41st Annual Meeting of the Association for Computational Linguistics, Sapporo, Japan, July 2003, 73-80. 
Professor of Computational Linguistics
University of Toronto, Department of Computer Science