topic Re: Is anyone using Decipher in non-alphabetic languages? in Product Forum

Is anyone using Decipher in non-alphabetic languages?

sumire — Thu, 15 Aug 2024 04:52:28 GMT

Hello,

If anyone uses Decipher in non-alphabetic language, please let me know what improvements you have made to OCR.
I am using Decipher for Japanese, but I am at a loss to learn that the accuracy of character recognition is poor and that Decipher cannot be trained for OCR.

Re: Is anyone using Decipher in non-alphabetic languages?

steven.boggs — Thu, 15 Aug 2024 16:22:45 GMT

Which version of Decipher IDP are you currently using?

The release of v2.3 increased the number of supported languages and currencies to include Japanese, Chinese, and Korean. With this release, Decipher IDP now supports more than 30 languages -- if you're not doing so already, have you tried using the latest release to see if this improves the character recognition?

Re: Is anyone using Decipher in non-alphabetic languages?

sumire — Fri, 16 Aug 2024 06:48:13 GMT

Hello @steven.boggs ,

I use v2.3, it can read Japanese.
However, as you know, there are several types of characters in Japanese, and to be specific, the accuracy of katakana is particularly poor. (ex. "ク", "グ" and "ケ")
In order to use Decipher for business purposes, this problem must be cleared.

Re: Is anyone using Decipher in non-alphabetic languages?

Sangjun — Tue, 24 Sep 2024 01:24:44 GMT

Hello, sumire sensei.

I am using Decipher ver 2.3 in Korean language. I feel a level of recognition of Korean Letters is good, but the recognition for numbers with Korean Letters is not good. Especially, when dealing with the scanned images.

I wonder there are some ways to enhance the recognition when handling numbers.

Re: Is anyone using Decipher in non-alphabetic languages?

sumire — Tue, 24 Sep 2024 01:47:15 GMT

Hello @Sangjun ,

I'm lucky to meet you who has a similar problem.
It seems like it's hard to recognize the Korean numbers, too.
(ex. 7 "칠" and 8 "팔")

If we list the characters that are hard to recognize in each language, will it help you to improve Decipher? @steven.boggs

Re: Is anyone using Decipher in non-alphabetic languages?

Sangjun — Tue, 24 Sep 2024 01:59:21 GMT

Dear @sumire

My mention means when below

"안녕하세요3780912" from scanned images or papers

IDP successfully capture "안녕하세요", but cannot perfectly capture "3780912" ,
it captures like "3ㄱ80912" .

So, I am testing many conditions of the scanned images when it comes to the resolution, image quality, colored or mono(blank and white image)

Re: Is anyone using Decipher in non-alphabetic languages?

sumire — Tue, 24 Sep 2024 08:40:43 GMT

Hello @Sangjun ,

Even in Japanese capturing, there are cases similar to the case you exemplified in Korean letters.
エーエフハチビーニ captures 上ー工フハナヒ""ー二
Decipher often mistakenly captures katakana into symbols or kanji.