General questions

Answered

How to disable self-correcting in recognition?

Our app is an education app, we are using MyScript to recognize words that end users wrote. 

It seems to be self-correcting the spelling, for example, if users wrote 'locotion', it will be recognized as 'location'.

According to Linguistic knowledge, 'An LK is not mandatory but not attaching one often results in a significant accuracy drop.', so I tried deleting en_US-lk-text.res file, but it did not make any change, the self-correcting is still working. 

Could you tell me how to disable self-correcting?


image



Best Answer

Dear Lyxaxs,


Thank you for contacting us and your question.


If I understand well, you will like to have some kind of  dictation application, e.g. if the writer writes "helo", you would not  expect "hello"?


Currently this is a  use-case our technology has not been designed for. Indeed, our  technology uses lexicons (list of words), that help recognize words ;  the drawback is that this results in what we call "lexical attraction",  which is just what I explained.


There  is not much to do for this ; one solution we suggest for such use-case  is to create a custom lexicon that would contain all the most common  writing mistakes faced (e.g. helo, helllo...), but we have no  significant feedback for this.


Best regards,


Olivier

1 Comment

Answer

Dear Lyxaxs,


Thank you for contacting us and your question.


If I understand well, you will like to have some kind of  dictation application, e.g. if the writer writes "helo", you would not  expect "hello"?


Currently this is a  use-case our technology has not been designed for. Indeed, our  technology uses lexicons (list of words), that help recognize words ;  the drawback is that this results in what we call "lexical attraction",  which is just what I explained.


There  is not much to do for this ; one solution we suggest for such use-case  is to create a custom lexicon that would contain all the most common  writing mistakes faced (e.g. helo, helllo...), but we have no  significant feedback for this.


Best regards,


Olivier