Patterns Bot Workers use Vision Object Recognition to recognize and read contents from Documents. Banking and Financial Documents have the following types of objects to be recognized:
1. Text Characters of any language (like English, Hindi, Arabic, etc.)
2. Non-Standard Objects used while writing Financial Amounts
3. Logos, Icons and any other graphical object to be recognized
4. Stamps and Endorsements
5. Human Signature
OCR cannot recognize Non-Standard Objects, Graphic Objects, Human Signature and Cursive handwritten characters. OCR needs format templating to be pre-defined to read from documents.
Patterns Vision Bot Document Reading technology is much Beyond OCR, due to the following reasons:
1. No prior templating of document is required
2. Text Characters of Multiple languages like English, Arabic, Japanese, etc. can be recognized
3. Non-Standard symbols handwritten can be recognized (mostly as de-limiters to amounts)
4. Graphic Icons, Objects, etc. can be recognized
5. Cursive handwritten text can be recognized, either with cursive intelligent breakup marking or with multi-character recognition
6. Multiple recognizers: CNN Model to read printed and handwritten text, RNN Model to recognize words while limited vocabulary is needed, Feature Recognizer based on Patterns Definition Language (PDL) feature description of reference objects, used when CNN / RNN models fail to recognize and Reviewer Bot to recheck the recognition and increase quality of recognition. Different Sets of Bots are made available for each Language Content to be processed.
7. On the Job and continuous learning using Teacher Bot Models, that use human repairs carried out for wrong recognitions by generating PDLs to be used by the Feature Recognizer.
8. Content based Document Classifier (Document Recognizer)
9. Content Based – Label Based Data Element (Field) Recognizer
10. Document Signature Presence and Signature Matching with specimen Signatures
Beyond OCR, simply means Bot Worker Vision Technology that provides a very powerful and sophisticated document processing capability in Hyper Automation Implementations.
Nov
27