Integrate Document AI

Document AI could be a useful addition to the STRR system when it comes to handling large volumes of uploaded documents of different types. The tool can be used to extract structured data such as identifying numbers on licences, expiration dates, etc.

There are around 3000+ registrations with business licence required and another 3000 coming soon.

This work will be valuable to examiners by extracting and presenting data in a more direct fashion, saving time from having to load a PDF file or image to inspect the contents.

A confidence level is set for each data element and overall for the document.

First pass could be to iterate through all documents, extract data in an excel file so that the data can be accessed in bulk
Second pass could be to enhance examiner UIs to include data points
Future work could further offload workflow

Questions:
- what data is important?
- how would examiners use this information?
- Would a confidence level be useful?
- Could we describe the examiners workflow in a way that could allow for automation?
- Could further automation be useful for future improvements?
- Can we get a sample document from the upcoming municipalities
- Could we use this for all acceptable documents - prioritize important documents first
- Can we use the existing set of documents provided in the registry to train the data model?
- Service is already running in the Canadian Cluster - Approval is already in place for use on STRR
- What does the long term maintenance picture look like?