bms-treatment

This repo contains the CTDI Treatment module. In it you will find how to build a pipeline and execute an NLP Pipeline with the following tasks on excerpts of natural language text:

• Clinical Word Embeddings
• Named Entity Recognition for Drugs and Treatments
  • Regex Matcher
  • Text Matcher
  • Deep Learning NER
• Assertion for Treatments (Past/Present/Absent...)
• RxNorm Entity Resolution for Drugs and their has_disposition relationships

It also contains a notebook to understand how to use the module for two particular use cases:

• Run the Pipeline on ARM and Detailed Intervention information from AACT Database and parsing it into a revised Treatment taxonomy while complementing the objects at Intervention level with the information at ARM level. The output of this use case is a csv with some columns in json.
• Run the Pipeline on ARM and Aggregated Intervention information from AACT Database and preannotate it in order to upload it to the Annotation Lab. The output of this use case is a json.

In the models folder you can adjust the behavior of the following models:

• RegexMatcher: Use the arms_regex.csv to add additional regex patterns
• TextMatcher: Use the arms_treatment_textmatcher.csv to add additional keyword paths
• NER Output: Use the fp_dict.csv to prevent the inclusion of certain known false positives