Disease specific biomedical literature mining : A visual interface

Abstract

"The vast amount of online resources available in the biomdedical domain makes it challenging for the user to fulfill an information need. Most of the existing approaches for biomedical text mining try to alleviate the problem of information abundance by information extraction, summarization, etc and present the output in text format. In this study we present an approach to extract potential research topics from available biomedical literature, and analyse them in a temporal and geographical framework. In order to fulfill this task we have utilized the UMLS resource and geotext library. We identify important milestones in the lifecycle of each research topic using an approach based on topic novelty and published volume. Closely related concepts are identified and represented as graphs. We have also designed a novel visual interface for representing the multifaceted information extracted by this approach. The experiments have been performed on MEDLINE citations accessible through PubMed. We evaluate the performance of our approach using the metric recision. The results of our approach are presented and scope of improvement identified."