ITR: Text Metadata Mining: Extending the Frontiers of Text Based Applications in Biomedicine

University Of Iowa, Iowa City IA

Investigators

Abstract

The goal in text mining is automatic knowledge discovery from large text collections. The importance of text mining technology comes from its potential to enable the process of hypothesis generation and thereby influence the progress of science. This complex process, a crucial initial step for making scientific discoveries, relies on several factors including intangible ones such as prior experience and intuition. Oftentimes chance connections made serendipitously, later turn out to be fruitful. Text mining tools automatically scour large text collections to identify a small set of hypotheses that are both novel and interesting enough to warrant further research by the user. This research is a study of text mining algorithms, especially those that exploit the metadata associated with text. The specific aims are (1) to implement algorithms designed around three basic text mining functions in a prototype hypothesis generation and knowledge discovery tool and (2) to evaluate this prototype through a set of mining experiments designed to explore knowledge discovery within the biomedical domain. Although biomedicine forms the topical context of this study, the mining functions explored are general and may be applied to any subject area that has a large text database with associated metadata.