Tools for protecting vital healthcare programs by extracting actionable knowledge from policy.
This code was originally used as part of a pipeline for Policy Knowledge Extraction, which aims to make a policy-compliance ‘landscape’ visible to healthcare programs - helping policy investigators identify Fraud, Waste or Abuse.
This pipeline ingests PDF policy documents and processes them with their metadata, transforming the documents into a hierarchical tree structure. It performs concept annotation, rule extraction and consolidation guided by reasoning and checking consistency over the domain ontology. The ontology guides the transformation of textual patterns into a semantically-meaningful knowledge graph of rules.
For more information on the pipeline, the ontology lifting and examples see:
Towards protecting vital healthcare programs by extracting actionable knowledge from policy. Vanessa Lopez, Nagesh Yadav, Gabriele Picco, Inge Vejsbjerg, Eoin Carroll, Seamus Brady, Marco Luca Sbodio, Lam Thanh Hoang, Miao Wei, John Segrave. Findings of ACL, 2021
Futher information on the domain ontology can be found in:
An base ontology is provided which can be supplemented by domain-specific external entities. Follow the instructions in the sections below to generate the ontology and access it though a REST service.
- Run the application
- Add external entities to the ontology
- Restart and rebuild ontology
- Access the REST API
Navigate to the folder rules_extraction_from_healthcare_policy/ontology-and-lifting/ and run the following docker command.
docker build --tag policy-knowledge-extraction-service . && docker run -p 8086:8086 policy-knowledge-extraction-service
The policy knowledge extraction system adds domain-specific external entities such as places of service, CPT and HCPCS codes to the ontology.
For information and full instructions about lifting external data to be used during the extraction process, refer to this document. This process is required, for example, to
load a new set of procedure codes in the domain ontology. Some predefined configuration files with corresponding csv
files are available in external-data/* folders. Some group mappings are provided as a proof-of concept (a small sample for evaluation purposes covering our dental policies benefit rules) to show the mapping from broad terms in policy text to specific procedure codes.
Stop the docker container and then rebuild your docker image:
docker build --tag policy-knowledge-extraction-service . && docker run -p 8086:8086 policy-knowledge-extraction-service
Verify that it is running by accessing http://localhost:8086/swagger-ui.html#/
Swagger API documentation should now be available at http://localhost:8086/swagger-ui.html#/
The ontology can be exported using the Swagger API at http://localhost:8086/swagger-ui.html#/ontology-controller/getOntologyUsingGET in JSONLD, N3, NTRIPLES, RDFXML, or TURTLE format.
You can use RDF4j for example to explore the updated ontology file:
docker run -p 8080:8080 ghcr.io/ibm/rdf4j-policy-extraction-ontology
RDF4J Workbench is available at http://localhost:8080/rdf4j-workbench and contains the populated ontology that can be obtained using the lifting endpoints described above.