Gooch, P. & Roudsari, A. (2011). Coreference resolution in clinical discharge summaries, progress notes, surgical and pathology reports: a unified lexical approach. Paper presented at the AMIA 2011, 22 - 26 Oct 2011, Washington DC, US.
Download (339kB) | Preview
We developed a lexical rule-based system that uses a unified approach to resolving coreference across a wide variety of clinical records comprising discharge summaries, progress notes, pathology, radiology and surgical reports from two corpora (Ontology Development and Information Extraction (ODIE) and i2b2/VA) provided for the fifth i2b2/VA shared task. Taking the unweighted mean between 4 coreference metrics, validation of the system against the i2b2/VA corpus attained an overall F-score of 87.7% across all mention classes, with a maximum of 93.1% for coreference of persons, and a minimum of 77.2% for coreference of tests. For the ODIE corpus the overall F-score across all mention classes was 79.4%, with a maximum of 82.0% for coreference of persons and a minimum of 13.1% for coreference of diagnostic reagents. For the ODIE corpus our results are comparable to the mean reported inter-annotator agreement with the gold standard. We discuss the four categories of errors we identified, and how these might be addressed. The system uses a number of reusable modules and techniques that may be of benefit to the research community.
|Item Type:||Conference or Workshop Item (Paper)|
|Subjects:||R Medicine > RA Public aspects of medicine|
Z Bibliography. Library Science. Information Resources > Z665 Library Science. Information Science
|Divisions:||School of Informatics > Centre for Health Informatics|
Actions (login required)