Narratives of electronic health records (EHRs) contain useful information that is difficult to automatically extract, index, search, or interpret. Clinical natural language processing (NLP) technologies for automatic extraction, indexing, searching, and interpretation of EHRs are in development; however, due to privacy concerns related to EHRs, such technologies are usually developed by teams that have privileged access to EHRs in a specific institution. Technologies that are tailored to a specific set of data from a given institution generate inspiring results on that data; however, they can fail to generalize to similar data from other institutions and even other departments from the same institution. Therefore, learning from these technologies and building on them becomes difficult. In order to improve NLP in EHRs, there is need for head-to-head comparison of approaches that can address a given task on the same data set. Shared-tasks provide one way of conducting systematic head-to- head comparisons. This proposal describes a series of shared-task challenges and conferences, spread over a five year period, that promote the development and evaluation of cutting edge clinical NLP systems by distributing de-identified EHRs to the broad research community, under data use agreements, so that: * the state-of-the-art in clinical NLP technologies can be identified and advanced, * a set of technologies that enable the use of the information contained in EHR narratives becomes available, and * the information from EHR narratives can be made more accessible, for example, for clinical and medical research. The scientific activities supporting the organization of the shared-task challenges are sponsored in part by Informatics for Integrating Biology and the Bedside (i2b2), grant number U54-LM008748, PI: Kohane. This proposal aims to organize a series of workshops, conference proceedings, and journal special issues that will accompany the shared-task challenges in order to disseminate the knowledge generated by the challenges.

Public Health Relevance

this proposal will address two main challenges related to the use of clinical narratives for research: availability of clinical records for research and identification of the state of the art in clinical natural language processing (NLP) technologies so that we can push the state of the art forward and so that future work can build on the past. Progress in clinical NLP will improve access to electronic health records for research, and for clinical applications, benefiting healthcare and public health.

National Institute of Health (NIH)
National Library of Medicine (NLM)
Conference (R13)
Project #
Application #
Study Section
Biomedical Library and Informatics Review Committee (BLR)
Program Officer
Sim, Hua-Chuan
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
State University of New York at Albany
Schools of Arts and Sciences
United States
Zip Code
Singh, Vivek Kumar; Shrivastava, Utkarsh; Bouayad, Lina et al. (2018) Machine learning for psychiatric patient triaging: an investigation of cascading classifiers. J Am Med Inform Assoc 25:1481-1487
Karystianis, George; Nevado, Alejo J; Kim, Chi-Hun et al. (2018) Automatic mining of symptom severity from psychiatric evaluation notes. Int J Methods Psychiatr Res 27:
Liu, Zengjian; Tang, Buzhou; Wang, Xiaolong et al. (2017) De-identification of clinical notes via recurrent neural network and conditional random field. J Biomed Inform 75S:S34-S42
Uzuner, Özlem; Stubbs, Amber; Filannino, Michele (2017) A natural language processing challenge for clinical records: Research Domains Criteria (RDoC) for psychiatry. J Biomed Inform 75S:S1-S3
Dehghan, Azad; Kovacevic, Aleksandar; Karystianis, George et al. (2017) Learning to identify Protected Health Information by integrating knowledge- and data-driven algorithms: A case study on psychiatric evaluation notes. J Biomed Inform 75S:S28-S33
Rios, Anthony; Kavuluru, Ramakanth (2017) Ordinal convolutional neural networks for predicting RDoC positive valence psychiatric symptom severity scores. J Biomed Inform 75S:S85-S93
Lee, Hee-Jin; Zhang, Yaoyun; Roberts, Kirk et al. (2017) Leveraging existing corpora for de-identification of psychiatric notes using domain adaptation. AMIA Annu Symp Proc 2017:1070-1079
Clark, Cheryl; Wellner, Ben; Davis, Rachel et al. (2017) Automatic classification of RDoC positive valence severity with a neural network. J Biomed Inform 75S:S120-S128
Stubbs, Amber; Filannino, Michele; Uzuner, Özlem (2017) De-identification of psychiatric intake records: Overview of 2016 CEGS N-GRID shared tasks Track 1. J Biomed Inform 75S:S4-S18
Lee, Hee-Jin; Wu, Yonghui; Zhang, Yaoyun et al. (2017) A hybrid approach to automatic de-identification of psychiatric notes. J Biomed Inform 75S:S19-S27

Showing the most recent 10 out of 77 publications