iink SDK on Device

Answered

Korean recognition

Hello.

I am making Korean dictation android app.

To score dictation, you must recognize it even if you misspell it.

But when I do editor.export_ it is replaced by the words in the dictionary.

How can I fix it?

 

You can also use English as Korean, but sometimes Chinese comes out suddenly. How can I remove Chinese?


My setup information.

--------------------------------------
Bundle-Version: 1.0

Bundle-Name: ko_KR

Configuration-Script:

 AddResDir ../resources/

 

Name: text

Type: Text

Configuration-Script:

 AddResource ko_KR/ko_KR-ak-cur.res

 AddResource ko_KR/ko_KR-lk-text.res

 AddResource mul/mul-lk-english_secondary_2.res

 EnableAlienCharacters

 SetTextListSize 1

 SetWordListSize 5

 SetCharListSize 1


Best Answer

Dear Gunoh Kim,


thank you for contacting us.


If I understand well, you would like to use our solution for a dictation application?


Currently, our solution is using lexicons, that help improve the accuracy. A consequence is that when an ink is not properly written, result is usually "corrected" ; this is called lexical attraction.


As you understand, this is not compatible with a dictation application. But oyu understand nothing can be done to prevent this.


Regarding the Chinese, if I understand well, you would like to disable the Hanja? For this, I just compile for you the following Subset Knowledge (SK resource): https://myscript.filecamp.com/s/THsJKcj7JsQx3U2K/fo


I recommend you add it in the "ko_KR" assets of your project. You shall then update your configuration as follows:

Bundle-Version: 1.0 

Bundle-Name: ko_KR 

Configuration-Script: 

 AddResDir ../resources/ 

 

Name: text 

Type: Text 

Configuration-Script: 

 AddResource ko_KR/ko_KR-ak-cur.res 

 AddResource ko_KR/ko_KR-lk-text.res 

 AddResource ko_KR/disable_hanja.res 

 AddResource mul/mul-lk-english_secondary_2.res 

 EnableAlienCharacters 

 SetTextListSize 1 

 SetWordListSize 5 

 SetCharListSize 1


Best regards,


Olivier


Dear Gunoh Kim,


I just added the "disable-hanja+emoji.res" SK file at the above link, which will allow you disable emojis and hanja characters.


You shall then proceed the same way, i.e. add the resource in the "ko_KR" assets of your projects, and set your .conf file accordingly:

Name: text 

Type: Text 

Configuration-Script: 

 AddResource ko_KR/ko_KR-ak-cur.res 

 AddResource ko_KR/ko_KR-lk-text.res 

AddResource ko_KR/disable-hanja+emoji.res 

 AddResource mul/mul-lk-english_secondary_2.res 

 EnableAlienCharacters 

 SetTextListSize 1 

 SetWordListSize 5 

 SetCharListSize 1



Best regards,


Olivier

Dear Gunoh Kim,


Thank you for your update. 


No you can't add two Subset Knowledge files.

So you have to build a single .sk file containing the set of characters you want.

For the structure of .sk file, please refer to:

https://developer.myscript.com/docs/interactive-ink/1.3/android/advanced/build-custom-resources/#subset-knowledge


Once you have written your .sk file you should use the online tool to compile your .sk file to generate your .res file.

The online tool is available on https://developer.myscript.com/support/recognition-assets

Please, make sure to select Subset knowledge (Text) resource radio button before droping your .sk file.


Best regards,


Gwenaëlle



Thank Olivier

Your answer was well understood.

 

We have successfully removed hanja using the disable_hanja.res file.

 

Additionally, I went around the forum and found and applied the no-emojis.res file.

However, it has been confirmed that it does not apply simultaneously with the disable_hanja.res file.

 

Is there a way to apply two files at the same time?

 

I tried to create and apply a .sk file myself, but I can't figure out the structure of the .res file.

 

Is there a way to revert a .res file to a .sk file?

Answer

Dear Gunoh Kim,


thank you for contacting us.


If I understand well, you would like to use our solution for a dictation application?


Currently, our solution is using lexicons, that help improve the accuracy. A consequence is that when an ink is not properly written, result is usually "corrected" ; this is called lexical attraction.


As you understand, this is not compatible with a dictation application. But oyu understand nothing can be done to prevent this.


Regarding the Chinese, if I understand well, you would like to disable the Hanja? For this, I just compile for you the following Subset Knowledge (SK resource): https://myscript.filecamp.com/s/THsJKcj7JsQx3U2K/fo


I recommend you add it in the "ko_KR" assets of your project. You shall then update your configuration as follows:

Bundle-Version: 1.0 

Bundle-Name: ko_KR 

Configuration-Script: 

 AddResDir ../resources/ 

 

Name: text 

Type: Text 

Configuration-Script: 

 AddResource ko_KR/ko_KR-ak-cur.res 

 AddResource ko_KR/ko_KR-lk-text.res 

 AddResource ko_KR/disable_hanja.res 

 AddResource mul/mul-lk-english_secondary_2.res 

 EnableAlienCharacters 

 SetTextListSize 1 

 SetWordListSize 5 

 SetCharListSize 1


Best regards,


Olivier


1 person likes this