RelAgent: Entity Detection and Normalization for Diseases in Clinical Records: a Linguistically Driven Approach

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 477–481, Dublin, Ireland, August 23-24, 2014. RelAgent: Entity Detection and Normalization for Diseases in Clinical Records: a Linguistically Driven Approach Sv Ramanan RelAgent Tech Pvt Ltd Adyar, Chennai India ramanan@relagent.com Senthil Nathan RelAgent Tech Pvt Ltd Adyar, Chennai India senthil@relagent.com Abstract We refined the performance of Co- coa/Peaberry, a linguistically moti- vated system, on extracting disease en- tities from clinical notes in the train- ing and development sets for Task 7. Entities were identified in noun chunks by use of dictionaries, and events (‘The left atrium is dilated’) through our own parser and predicate-argument struc- tures. We also developed a mod- ule to map the extracted entities to the SNOMED subset of UMLS. The module is based on direct matching against UMLS entries through regu- lar expressions derived from a small set of morphological transformations, along with priority rules when multi- ple UMLS entries were matched. The performance on training and develop- ment sets was 81.0% and 83.3% respec- tively (Task A), and the UMLS match- ing scores were respectively 75.3% and 78.2% (Task B). However, the perfor- mance against the test set was low by comparison, 72.0% for Task A and 63.9% for Task B, even while the pure UMLS mapping score was reasonably high (relaxed score in Task B = 91.2%). We speculate that our moderate perfor- mance on the test set derives primarily from chunking/parsing errors. 1 Introduction The increasing use of electronic health records, both for satisfying mandatory requirements as This work is licensed under a Creative Commons At- tribution 4.0 International Licence. Page numbers and proceedings footer are added by the organisers. Licence details: http://creativecommons.org/licenses/by/ 4.0/ well as for administrative reasons, has cre- ated a need for systems to automatically tag and normalize disease/sign/symptom men- tions. Statistically significant correlations extracted from automated analysis of large databases of clinical records are felt to be use- ful in detecting phenotype-genotype correla- tions (reviewed in Kohane (2011)), phenotype- phenotype correlations (Roque et al., 2011) as well as in continuous monitoring of events such as adverse reactions and even early detection of outbreaks of epidemics/infectious diseases (Botsis et al., 2013; Collier, 2012). In this context, Task 7 of SemEval 2014, which is a continuation of the ShARe/CLEF eHealth 2013 task (Pradhan et al., 2013), provides a testbed to evaluate systems that automatically tag and normalize mentions of diseases, signs and symptoms in clinical records, which in- clude discharge summaries and echo, radiology and ECG reports. Our system consists of (i) Cocoa, a chunk- based entity tagger and (ii) Peaberry, a parser, followed by a module for predicate-argument structure. We have tested the system in a va- riety of tasks, such as detecting and normal- izing mentions of chemicals, proteins/genes, diseases and action terms in the BioCreative 13 Chemdner and CTD tasks (Ramanan and Senthil Nathan, 2013a; Ramanan and Senthil Nathan, 2013b), as well as in detecting cel- lular and pathological events in the BioNLP cancer genetics task …

Content truncated to 500 words