IHS-RD-Belarus: Identification and Normalization of Disorder Concepts in Clinical Notes

Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015), pages 380–384, Denver, Colorado, June 4-5, 2015. c⃝2015 Association for Computational Linguistics IHS-RD-Belarus: Identification and Normalization of Disorder Concepts in Clinical Notes Maryna Chernyshevich Vadim Stankevitch IHS Inc. / IHS Global Belarus IHS Inc. / IHS Global Belarus 131 Starovilenskaya St 131 Starovilenskaya St 220123, Minsk, Belarus 220123, Minsk, Belarus {Marina.Chernyshevich}@ihs.com {Vadim.Stankevitch}@ihs.com Abstract This paper describes clinical disorder recognition and encoding system submit- ted by IHS R&D Belarus team at the SemEval-2015 shared task related to analysis of clinical texts. Our system is based on IHS Goldfire Linguistic Proces- sor and uses a rich set of lexical, syntac- tic and semantic features. The proposed system consists of two components: a CRF-based approach to recognize disor- der entities and empirical ranking to en- code disorders to UMLS CUIs. Evalua- tion on the test data set showed that our system achieved the F-measure of 0.898 for entity recognition and the F-measure of 0.794 for UMLS CUI. The combined score for whole task is 0.690 (rank 17 out of 40 submissions). 1 Introduction Named entity recognition (NER) is an infor- mation extraction task where the aim is to identi- fy mentions of specific types of entities in text. This task has been one of the main focuses in the biomedical text mining research field, especially when applied to the scientific literature. Such efforts have led to the development of various tools for the recognition of diverse entities, in- cluding species names, genes and proteins, chemicals and drugs, anatomical concepts and diseases. These tools use methods based on dic- tionaries, rules, and machine learning, or a com- bination of those depending on the specificities and requirements of each concept type (Campos et al., 2013). After identifying entities occurring in texts, it is also relevant to disambiguate those entities and associate each occurrence with a specific concept, using a univocal identifier from a reference database such as Uniprot1 for pro- teins, or OMIM2 for genetic disorders. This is usually performed by matching the identified entities against a knowledge-base, possibly evaluating the textual context in which the entity occurred to identify the best matching concept. In this paper, we describe a system (IHS_RD_Belarus in official results) developed to participate in the international shared task organized by the Conference on Semantic Evalu- ation Exercises (SemEval-2015) and focused on the analysis of clinical notes. This task is the repetition of task 7 at SemEval-2014 (Pradhan, et al., 2014) and aims at the recognition of entities belonging to the ‘disorders’ semantic group of the Unified Medical Language System (UMLS) (Bodenreider, 2004) and normalization of these entities to a specific UMLS Concept Unique Identifier (CUI). Specifically, the task definition required that concepts should only be normalized to CUIs that could be mapped to the SNOMED CT3 terminology. 2 System description 2.1 Dataset The dataset for Tasks 1 consists of de-identified clinical notes of 4 different types (Discharge summary, ECG, Echo, Radiology) from MIMIC corpus (Lee et al., 2011). The …

Content truncated to 500 words