UTA DLNLP at SemEval-2016 Task 12: Deep Learning Based Natural Language Processing System for Clinical Information Identification from Clinical Notes and Pathology Reports

Proceedings of SemEval-2016, pages 1268–1273, San Diego, California, June 16-17, 2016. c⃝2016 Association for Computational Linguistics UTA DLNLP at SemEval-2016 Task 12: Deep Learning Based Natural Language Processing System for Clinical Information Identification from Clinical Notes and Pathology Reports Peng Li Computer Science and Engineering University of Texas at Arlington jerryli1981@gmail.com Heng Huang∗ Computer Science and Engineering University of Texas at Arlington heng@uta.edu Abstract We propose a deep neural network based nat- ural language processing system for clinical information (such as time information, event spans, and their attributes) extraction from raw clinical notes and pathology reports. Our ap- proach uses the context words and their part- of-speech tags and shape information as fea- tures. We utilize the temporal (1D) convolu- tion neural network to learn the hidden fea- ture representations. In prediction step, we use the Multilayer Perceptron (MLP) to predict event spans. The empirical evaluation demon- strates that our approach significantly outper- forms baseline methods. 1 Introduction In past several years, there has been much interest in applying neural network based deep learning tech- niques to solve many natural language processing (NLP) tasks. From low-level tasks such as language modeling, POS tagging, named entity recognition, and semantic role labeling (Collobert et al., 2011; Mikolov et al., 2013), to high-level tasks such as machine translation, information retrieval, semantic analysis (Kalchbrenner and Blunsom, 2013; Socher et al., 2011a; Tai et al., 2015) and sentence relation modeling tasks such as paraphrase identification and question answering (Socher et al., 2011b; Iyyer et al., 2014; Yin and Schutze, 2015). Deep represen- tation learning has demonstrated its importance for ∗To whom all correspondence should be addressed. This work was partially supported by NSF-IIS 1117965, NSF-IIS 1302675, NSF-IIS 1344152, NSF-DBI 1356628, NIH R01 AG049371. these tasks. All the tasks get performance improve- ment via learning either word level representations or sentence level representations. In this work, we introduce the deep representa- tion learning technologies to the electronic medi- cal record research. Specifically, we focus on clin- ical information extraction, using clinical notes and pathology reports from the Mayo Clinic. Our system is designed to identify event expressions consisting of the following components: • The spans (character offsets) of the expression in the raw text • Contextual Modality: ACTUAL, HYPOTHET- ICAL, HEDGED or GENERIC • Degree: MOST, LITTLE or N/A • Polarity: POS or NEG • Type: ASPECTUAL, EVIDENTIAL or N/A The input of our system consists of raw clinical notes or pathology reports. The following is an ex- ample: April 23, 2014: The patient did not have any postoperative bleeding so we will resume chemotherapy with a larger bolus on Friday even if there is slight nausea. The output annotations over the text capture the key information such as event mentions and at- tributes. Table 1 illustrates the output of clinical in- formation extraction in details. To solve this task, the major challenge is how to precisely identify the spans (character offsets) of 1268 Clinical Note Event Mention Event Attribute April 23, 2014: The patient did …

Content truncated to 500 words