UTH_CCB: A report for SemEval 2014 – Task 7 Analysis of Clinical Text

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 802–806, Dublin, Ireland, August 23-24, 2014. UTH_CCB: A Report for SemEval 2014 – Task 7 Analysis of Clinical Text Yaoyun Zhang1 Jingqi Wang1 Buzhou Tang2 Yonghui Wu1 Min Jiang1 Yukun Chen3 Hua Xu1* 1University of Texas School of Biomedical Informatics at Houston Houston, TX, 77030, USA 2Harbin Institute of Technology Shenzhen Graduate School Shenzhen, 518055, China 3Vanderbilt University Department of Biomedical Informatics Nashville, TN, 37240, USA {Yaoyun.Zhang, Yonghui.Wu, Min.Jiang, Hua.Xu} @uth.tmc.edu tangbuzhou@gmail.com yukun.chen@Vanderbilt.Edu Abstract This work describes the participation of the University of Texas Health Science Center at Houston (UTHealth) team on the SemEval 2014 – Task 7 analysis of clinical text challenge. The task consisted of two subtasks: (1) disorder entity recognition, recognizing mentions of disorder concepts; (2) disorder entity encoding, mapping each mention to a unique Concept Unique Identifier (CUI) defined in Unified Medical Language System (UMLS). We developed three ensemble learning approaches for recognizing disorder entities and a Vector Space Model based method for encoding. Our approaches achieved top rank in both subtasks, with the best F measure of 0.813 for entity recognition and the best accuracy of 74.1% for encoding, indicating the proposed approaches are promising. 1 Introduction In recent years, clinical natural language processing (NLP) has received great attention for its critical role in unlocking information embedded in clinical documents. Leveraging such information can facilitate the secondary1 use of electronic health record (EHR) data to This work is licensed under a Creative Commons Attribution 4.0 International Licence. Page numbers and proceedings footer are added by the organisers. Licence details:http://creativecommons.org/licenses/by/4.0/ promote clinical and translational research. Clinical entity recognition, which recognizes mentions of clinically relevant concepts (e.g., disorders, procedures, drugs etc.) in narratives, and clinical entity encoding, which maps the recognized entities to concepts in standard vocabularies (e.g., UMLS CUI (Bodenreider, 2004)), are among the fundamental tasks in clinical NLP research. Many systems have been developed to extract clinical concepts from various types of clinical notes in last two decades, ranging from early symbolic NLP systems heavily dependent on domain knowledge to machine learning algorithm based systems driven by increasingly available annotated clinical corpora. The representative systems include MedLEE (Friedman et al., 1994), MetaMap (Aronson and Lang, 2010), KnowledgeMap (Denny et al., 2003), cTAKES (Savova et al., 2010), etc. Clinical NLP challenges organized by the Center for Informatics for Integrating Biology & the Beside (i2b2) have promoted research using machine learning algorithms to recognize clinical entities (Uzuner et al., 2010; Uzuner et al., 2011). Unlike the previous i2b2 challenges, the ShARe/CLEF challenge of clinical disorder extraction and encoding held in 2013 took the initiative to recognize disjoint entities, in addition to entities made up of consecutive words (Chapman et al., 2013). ShARe/CLEF challenge also required encoding of the disorder entities to Systematized Nomenclature Of Medicine Clinical Terms (SNOMED-CT) (using UMLS CUIs). 802 In this paper, we describe our system for Task 7 of SemEval 2014, which followed the requirements of 2013 ShARe/CLEF challenge. Our system …

Content truncated to 500 words