The Receipt Skill is a fully trained Vantage Document Skill which extracts key data points from receipts for a variety of expense types such as general retail, restaurant, car parking, transport tickets, etc. The skill has been fully trained on a large document set from the supported countries and provides exceptionally high accuracy of data identification and extraction. The Receipt Skill cannot be further trained on customer documents.
- Automatically applies image enhancement to images of receipts captured with a mobile phone in order to ensure optimal OCR results.
- Automatically detects receipt country and expense type. Fully trained for the expense types general retail, restaurant, gasoline station.
- Pre-trained model automatically detects key fields on the receipt and extracts relevant data (date, total, currency, vendor, taxes etc.). For a full list of extracted fields, please see the skill documentation.
- Includes business rules for required fields and to calculate total tax amount. Further business rules can be added as individually required.
- Can be augmented by human-in-the-loop review and verification based on accuracy thresholds, missing or incorrect information.
- Does not allow further training on customer documents or improvement of the existing model based on manual verification.
- The zip file includes 2 separate Document Skills: Receipt.zip (processes European Languages) and Receipt CJK.zip (processes Chinese, Japanese and Korean).
SarahSep 28, 2021