We are expanding some of our long term features to have new capabilities. In doing so, need to update their names for a more accurate description.
- The Custom Invoice Extractor has been changed to Semi-structured Document Extractor.
- The Docker function has been changed to Containerized function.
New Function: Tuple to Field
The Tuple to field function allows user to replace tuples with nested values with an element within the tuple.
This function is useful when trying to quickly use a single value from a nested tuple.
Tuple to field example
You are using an Invoice Extraction AI Model which generates tuple values.
Each OCRed field has four different sub-values. (Value, Snippit, Page Bounding Box, and Confidence Score).
If you quickly want to isolate only the confidence score of that field, this function can replace the entire tuple with the confidence score value.
You now have the confidence score value as a float data type that can be used within your Flow.
Function Improvement: IsLessThan
The IsLessThan formula function and the IsLessThan filter function have been updated to that users can choose between selecting field values from an input field or entering a static value.
New Knowledge Graph Feature: Match Strategy
The Fixed Form Document Extractor can use a Knowledge Graph to enhance the accuracy of values read in documents. A new feature, Search strategy, enhances the accuracy even further by letting the model know if it is looking for a single word or a phrase.
Relative vs Absolute Coordinates for Value Boxes in Semi-Structured Document Extraction
Before, only absolute coordinates were available for value box positioning in Invoice and Semi-structured Extraction models. This could cause problems when scaling documents so that the box position no longer matched the data it was set to capture.
Now, relative coordinates have been added for Invoice and Semi-structured models so the position of the value box always stays in the same position even when the document has scaled.