New to NLP

Knowledge Author (KA) is the first tool in a pipeline of tools from the IE-Viz (Information Extraction-Visualization) suite that aim to help users - clinicians, researchers, etc. - define, annotate, extract, assess, and visualize variables they would like to study. KA helps users define a schema or collection of variables they would like to study. These variables (e.g., fever, productive cough) and their contexts (e.g., whether its negated or affirmed, experienced by a patient or another, etc) - can be defined using the KA interface and exported for downstream extraction by additional tools in the IE-Viz suite. KA enables users to define variables using terminologies such as the Unified Medical Language System to ensure the models that are generated are interoperable with standard vocabularies and, thereby, other clinical natural language processing solutions.

System development was funded by VA CREATE funding; Ontologies were funding by SHARP, the Vårdal Foundation, the Interlock Project, the Stockholm University Academic Initiative, the National Library of Medicine, National Institute of General Medical Sciences, and National Institute of Health.

Know NLP

Knowledge Author (KA) is the first tool in a pipeline of tools from the IE-Viz (Information Extraction-Visualization) suite that aim to help users - clinicians, researchers, etc. - define variables, annotate, extract, assess, and visualize variables they would like to study. KA helps users define a schema or collection of variables they would like to study. These variables (e.g., fever, productive cough) and their contexts (e.g., whether its negated or affirmed, experienced by a patient or another, etc) - can be defined using the KA interface and exported for downstream extraction by additional tools in the IE-Viz suite. KA enables users to define variables using terminologies such as the Unified Medical Language System to ensure the models that are generated are interoperable with standard vocabularies and, thereby, other clinical natural language processing solutions. KA is comprised of two main representation ontologies: a schema ontology that defines the kinds of clinical elements e.g., problems, medications, and other elements supported and a modifier ontology that defines the types of contexts or properties each element can be described by e.g. both problems and medications can have experiencers, negation, and temporality. However, problems also have course and severity whereas medications have dose and route. These ontologies were created to be compatible with the CEM (Clinical Element Models), the SHARP project, the ShARe project, and previous schemas published by the BLULab. To instantiate a schema for a particular domain, a user defines a domain ontology by defining types of variables and their contexts based on the underlying property constraints of the schema and modifier ontologies. Upon completion, the schema can be exported into OWL format for further processing in the NLP pipeline.

Authors

Bill Scuba

Melissa Tharp

Wendy Chapman

Eugene Tseytlin

Yang Liu

Frank Drews

Associated Institutions

University of Utah

University of Pittsburgh

Minimum Requirements

Java

Download

Resource URLs

https://blulab.chpc.utah.edu/KA/

NLP Task Performed

Knowledge Base Dev

Programming Languages

Java

Operating Systems

WindowsMacLinux

Related Publications

Scuba W, Tharp M, Tseytlin E, Liu Y, Drews FA, Chapman WW. Knowledge Author: Creating Domain Content for NLP Information Extraction. In: Sixth International Symposium on Semantic Mining in Biomedicine (SMBM 2014). 2014. 99-103.