I'm currently testing OCR for refugee application documents using ABBYY Document AI API, but I'm not getting the desired results. The refugee application forms have a standardized layout with instructions in both Japanese and English. However, these documents are filled out by applicants who are often neither native Japanese nor English speakers, resulting in handwriting that includes spelling and grammatical errors. Additionally, applicants sometimes write outside the designated fields or extend their handwriting into the margins.
Given these challenges, is it possible to create a custom OCR model tailored specifically to handle these kinds of irregularities? If helpful, I can provide sample images that do not contain personal information for your evaluation. However, I would appreciate your careful handling of these samples.
Comments
1 comment
Hi Hiroshi, apologies for the delay! Given your requirements, you may need a custom OCR model which is something that ABBYY Vantage supports. I'm having someone from our team reach out soon with more thoughts - please stay tuned.
Please sign in to leave a comment.