Project Synopsis
A fast growing financial solution provider from the US wanted to roll out a functionality for their end users to scan receipts for expense settlement. The user could click a photograph of a receipt to be expensed using his phone camera and upload it for processing. The system was expected to extract the date, place, vendor, currency and the overall amount of the transaction from the receipt. We engineered a system that could process a large volume of uploaded receipts and use our data extraction engine to carry out rule based extraction of the above mentioned information.
The developed platform was expected to have the following features:
- Text recognition of information from receipts
- Extraction of currency, expense type, location, date and total value of transaction
- Automatic validation of the expenses against the submitted expense claim.
Rare Mile Solution
We designed and developed the complete solution using our text analytics assets and IP. The delivered solution had the following features:
- Support receipts for cabs, hotels, parking charges, retail store purchases, etc.
- Capability to automatically split multiple receipts scanned together in a single picture into individual receipts.
- OCR recognition and validation of extracted text
- Algorithms for image enhancements and clean up before processing
Project Highlights
- Created an ability to recognize non conclusive OCR scan by augmenting them with statistical model
- Developed a mobile application to capture shake free, sharply focused images
- Created new algorithms for processing receipts which could not be read by the OCR engine in the first pass
