Empirical data on corpus design and usage in biomedical natural language processing

This page provides a link to, and supplemental material related for, our 2005 AMIA paper. The paper provides data on the usage of various biomedical corpora. It discusses the formats of a couple of corpora that have not been frequently used outside of the labs that built them, and suggests that the format in which annotations are recorded and distributed is an important factor in determining whether or not a corpus will be widely adopted. At the suggestion of a reviewer, we performed a survey designed to elicit user preferences regarding various aspects of corpus design and contents. To see the survey and results, follow this link.

• Cohen, K. Bretonnel; Lynne Fox; Philip V. Ogren; and Lawrence Hunter (2005) Empirical data on corpus design and usage in biomedical natural language processing. Proceedings of AMIA 2005, pp. 156-160.
• Survey and results
• How we did the word counts: A symptom of the wide diversity of formats in which biomedical corpora are currently distributed is that we had to write code to count words in each corpus separately. This page gives information what we counted for each one.