Recommendations for interpreting Asian invoices

Due to the challenges that Asian characters present during interpretation, this topic provides specific recommendations that can be used to improve it.

Scanner settings

  • Use an image resolution of at least 400 DPI to get good recognition of CJK characters during interpretation when scanning images that contain Chinese, Japanese, or Korean characters.
  • Chinese invoices can be scanned in either color or black and white. Color images are larger and take more time to interpret, but they provide better character recognition. In addition, the data is easier to read manually on a color image.
  • Use Auto crop or Remove black borders scanner settings whenever possible so that the size of the image is the same as the size on the invoice when black/white borders are not visible.
  • If your scanner can rotate images automatically, enable it. However, note that the OrientationDetection flag in the [Interpret] section of eiglobal.ini cannot be used to automatically rotate images for Chinese (both simplified and traditional), Hebrew, Japanese, Korean, or Thai invoices.

Recommendations for automatic supplier identification

For Chinese invoices, we recommend using supplier master data that contains the VAT registration number. The interpretation of numerical values is significantly better than Chinese characters.

If master data is used, note that you can select which columns that are searchable in Verify, and make sure VAT registration number is selected. You can then either draw around the VAT registration number or manually type in the number in the supplier field to automatically identify the supplier. The results will not be as good if you draw around the supplier name since Chinese characters are not recognized as easily as numbers. You can also type the Chinese supplier name, of course.

Ctrl + draw

If you do not get the correct results when drawing around a field value in Verify, try using Ctrl + draw. However, note that it can only interpret numerical values and latin characters.