Enhanced entity extraction model with additional attributes support for unstructured text

This release introduces advanced capabilities for extracting a broader set of adverse event (AE) attributes from unstructured text sources such as narratives included in PDFs or emails.

Note:

AI-powered automatic extraction from source documents is for evaluation purposes only. Please review the output thoroughly before use.

• The system now supports extraction of an additional 17 fields for a total of 26 fields from unstructured text. An additional 8 relationships can also be extracted from unstructured text, totaling 13 relationships. This covers all relevant attributes such as Seriousness, Report Type, Initial Receipt Date, Lab Data, Patient Medical History, and expanded Patient, Product, and Event details.

  For a complete list of supported attributes, refer to Safety One Intake Unstructured Text Extraction List in the eTRM documentation.

• The underlying AI/ML models have been updated to support these additional attributes and further improve extraction accuracy. The model card can be provided upon request.