i2b2 Workshop on Natural Language Processing Challenges for Clinical Records

Abstract: This workshop aims to bring together computational linguists and medical informaticians interested in automatic linguistic processing of clinical records such as medical discharge summaries and radiology reports. Lack of a publicly available and standardized data set has been one of the biggest barriers to systematic progress of Natural Language Processing techniques for clinical data. Within the framework of the i2b2 project, we have generated and released a set of fully de-identified medical discharge summaries to the research community. We prepared two grand challenge questions around this data: What are some methods for automatic de-identification of clinical records and how well do they perform? Can we automatically identify the smoking status of patients based on their clinical records? We prepared the gold standard for both of these grand challenge questions. We made the data set available to interested researchers and invited them to participate in the grand challenge. At the time of writing, 18 teams had committed to participate. This workshop serves as a venue for the participants of the grand challenge to demonstrate their systems, present their papers, and discuss future directions i