Sergei Ananyan of Megaputer Intelligence makes a joint presentation with two Merck researchers on a PolyAnalyst text mining project at the Merck Technology Symposium 2008. Megaputer exhibits at IDEA (Booth EX21, Tuesday-Wednesday).
- Paul Kallukaran, Ph.D., Merck
- Razek Karnoub, Ph.D., Merck
- Sergei Ananyan, Ph.D., Megaputer Intelligence
In predicting the evolution of product sales volumes, one would benefit from knowing to what extent different clinical studies affect products and market performance. The potential impact of a clinical study can be determined through the analysis of research articles and other information discussing the study. However, direct manual analysis of large volumes of related literature requires major time and effort investments from a highly trained professional. Thorough analysis of all historical literature related to a particular issue might take several months of the analyst’s work.
We incorporated advanced text mining techniques to automate the initial screening of literature related to a large set of clinical studies on osteoporosis. Our objective was to automatically determine most likely impactful clinical studies and focus manual analysis efforts on literature discussing these impactful studies. The analysis was performed for 489 clinical studies related to osteoporosis. The list of likely impactful clinical studies identified through text mining was verified by a medical expert who performed prior manual analysis of the entire collection of articles related to all considered clinical studies. The studies identified through text mining were confirmed to be good candidates for further thorough manual investigation. The project demonstrated strong potential of text mining to dramatically reduce the amount of time domain experts would need to spend on determining potential impact of clinical studies on product performance.