The presented paper discusses a hybrid approach for negation processing in Electronic Health Records (EHRs) in Bulgarian. The rich temporal structure and the specific combination of medical terminology in both Bulgarian and Latin do not allow the application of standard language processing techniques. The problem gets even worse due to the often use of specific abbreviations, analyses and clinical test data. Various expressions of negation often occur in EHRs. This raises many difficulties for language processing especially in semantic analysis. That is why we propose an approach that combines information extraction with deep semantic analysis, allows to determine the negation, negation scope and to treat it appropriately. We present a prototype of a system MEHR for automatic recognition of medical terms and some facts in EHRs in Bulgarian. This is the first step towards filling in a template concerning the patient status. Automatic extraction of facts needed for description of patient status in full is our ultimate goal. However in this paper we just focus on proper treatment of negation.
[1]
George Hripcsak,et al.
An evaluation of natural language processing methodologies
,
1998,
AMIA.
[2]
Hsinchun Chen,et al.
A shallow parser based on closed-class words to capture relations in biomedical text
,
2003,
J. Biomed. Informatics.
[3]
Wendy W. Chapman,et al.
A Simple Algorithm for Identifying Negated Findings and Diseases in Discharge Summaries
,
2001,
J. Biomed. Informatics.
[4]
Ilya M. Goldin,et al.
Learning to Detect Negation with ‘Not’ in Medical Texts
,
2003
.
[5]
G Hripcsak,et al.
Natural language processing and its future in medicine.
,
1999,
Academic medicine : journal of the Association of American Medical Colleges.