Evaluating and improving annotation tools for medical forms
Abstract
The annotation of entities with concepts from standardized terminologies and ontologies is of high importance in the life sciences to enhance semantic interoperability, information retrieval and meta analysis. Unfortunately, medical documents such as clinical forms or electronic health records are still rarely annotated despite the availability of some tools to automatically determine possible annotations. In this study, we comparatively evaluate the quality of two such tools, cTAKES and MetaMap, as well as of a recently proposed annotation approach from our group for annotating medical forms. We also investigate how to improve the match quality of the tools by post-filtering computed annotations as well as by combining several annotation approaches.