Multimedia search over distributed collections

Our research will provide novel techniques for personalizing search results, allowing access to distributed, hetereogeneous resources We will use information from the patient record, query context, and user background to search and link to appropriate sources, translating the query into formats accepted by distributed, heterogeneous resources. Our research will involve automated indexing and classification of distributed resources according to source type, quality, and reading level to determine if a given document is appropriate for the specific end user. Results from the search will be filtered using natural language processing, rating those documents that match the patient record and user needs higher. To date, our results provide a means for determining the category of a remote textual source, for extracting terms from a text that can be used to match a patient record, and for segmenting and summarizing a particular type of multimedia, echocardiograms.