a week ago
Hi, I have a question. The original version of Decipher couldn’t support Thai, but the new version can uses Azure OCR Service as its core for OCR. Since Azure OCR supports Thai, does this mean Decipher with Azure OCR also supports Thai, or is there more to it?
a week ago
Changing the topic a bit but bringing an alternative, have you ever used Azure Form Recognizer? You can train the files and send them via API through Blue Prism. It sends the file, reads the mapped fields, and returns the fields via API. I have implemented this send and receive logic through BP and I still need to deploy it on DX.
Monday - last edited Monday
Hi,
Thanks for the earlier information! I’ve been working with Azure Document Intelligence for a while before exploring Decipher x Azure OCR. With Azure Document Intelligence, I could define document types and send documents through the API to get results. However, I faced a challenge: there’s no system or open-source example that lets business users interact with the processed documents (do you know of one?).
For instance, if some fields don’t meet accuracy thresholds, users should be able to open those documents, verify the data, and make annotations. The system should then collect this feedback for future training. At the moment, it feels like it’s just about sending and receiving results via the API, without any user interaction.
Recently, I found that Decipher’s latest update can connect to Azure services, like Document Intelligence and Azure OCR. Originally, Decipher didn’t support Thai, but since Azure OCR can read Thai, I tried connecting the two. While the connection to Azure OCR works, Decipher itself can’t read Thai. This makes me think that even though Azure OCR supports Thai, Decipher might not fully utilize that capability (perhaps due to how it processes text, switching from Tesseract to Azure OCR).
Next, I plan to test Decipher with Azure Document Intelligence to see if it can process documents and also offer features like data verification and classification verification. These would enable users to interact with documents directly.
This was a bit long, but I wanted to share my experience. Have you come across similar challenges or solutions in your work?