Activities

We have been working on two approaches to summarization, one for journal articles that would be appropriate for physician queries and the other for consumer health information that would be appropriate for patient queries.

For consumer health documents, which are not as clearly structured as are journal articles, we have been working on hierarchical segmentation of documents, using both layout cues and lexical chaining.

Also, as part of the analysis of lay texts, we have initiated a project to extract terms and their definitions from lay texts. The purpose of this research is to automatically populate and enhance existing technical and popular medical dictionaries and glossaries with terms and definitions extracted from on-line texts written at the lay level. Our output can be used in several ways: (1) to extend an existing dictionary; (2) to fill in gaps; (3) to be used in conjunction with summarization for presenting additional technical information to lay users in terms they can understand; (4) for text categorization to determine level of input text (e.g. technical, popular, etc.)