Fine tunning recognition for handwritten python source code recognition.
R
Rkodippily
started a topic
almost 2 years ago
My aim is to build an android app to convert handwritten stylus based source code into typset. Is it possible to fine tune the recognition to handle use cases like
1) correctly identifying camelCase variables
2) identifying variable_names_with_underscore
Any other general advice on how to increase accuracy when dealing with handwritten python source code would be greatly appreciated.
Thanks in advance!
Best Answer
O
Olivier @MyScript
said
almost 2 years ago
Dear Rkodippily,
thank you for your questions.
If I understand well, your use-case is that python code is hand-written and you would like to have it properly recognized?
Our technology has not been designed for such usage ; in particular, camel case is not supported in the iink SDK. In the same way, words terms with underscore are not supported.
For the camel case, you could try creating lexicons will all the variable/function names you would expect, writing these exactly with the casing you expect.
And for the word terms with underscore, you shall create a ludef/regular expression. The problem is that based on our experience, variables and function names can have any wording, and as our technololgy has been designed to recognize notes, we use lexicons that contain the most frequent words/first-names/last-names of a given language.
You then understand our technology is not much appropriate for such usage.
If I understand well, your use-case is that python code is hand-written and you would like to have it properly recognized?
Our technology has not been designed for such usage ; in particular, camel case is not supported in the iink SDK. In the same way, words terms with underscore are not supported.
For the camel case, you could try creating lexicons will all the variable/function names you would expect, writing these exactly with the casing you expect.
And for the word terms with underscore, you shall create a ludef/regular expression. The problem is that based on our experience, variables and function names can have any wording, and as our technololgy has been designed to recognize notes, we use lexicons that contain the most frequent words/first-names/last-names of a given language.
You then understand our technology is not much appropriate for such usage.
Best regards,
Olivier
R
Rkodippily
said
almost 2 years ago
Yes, my use case is to recognize handwritten python source code. Hard coding variable names into a custom lexicon is not really an option as you said since variable names can be anything. Thanks for the quick response!
Rkodippily
My aim is to build an android app to convert handwritten stylus based source code into typset. Is it possible to fine tune the recognition to handle use cases like
1) correctly identifying camelCase variables
2) identifying variable_names_with_underscore
Any other general advice on how to increase accuracy when dealing with handwritten python source code would be greatly appreciated.
Thanks in advance!
Dear Rkodippily,
thank you for your questions.
If I understand well, your use-case is that python code is hand-written and you would like to have it properly recognized?
Our technology has not been designed for such usage ; in particular, camel case is not supported in the iink SDK. In the same way, words terms with underscore are not supported.
For the camel case, you could try creating lexicons will all the variable/function names you would expect, writing these exactly with the casing you expect.
And for the word terms with underscore, you shall create a ludef/regular expression. The problem is that based on our experience, variables and function names can have any wording, and as our technololgy has been designed to recognize notes, we use lexicons that contain the most frequent words/first-names/last-names of a given language.
You then understand our technology is not much appropriate for such usage.
Best regards,
Olivier
- Oldest First
- Popular
- Newest First
Sorted by Oldest FirstOlivier @MyScript
Dear Rkodippily,
thank you for your questions.
If I understand well, your use-case is that python code is hand-written and you would like to have it properly recognized?
Our technology has not been designed for such usage ; in particular, camel case is not supported in the iink SDK. In the same way, words terms with underscore are not supported.
For the camel case, you could try creating lexicons will all the variable/function names you would expect, writing these exactly with the casing you expect.
And for the word terms with underscore, you shall create a ludef/regular expression. The problem is that based on our experience, variables and function names can have any wording, and as our technololgy has been designed to recognize notes, we use lexicons that contain the most frequent words/first-names/last-names of a given language.
You then understand our technology is not much appropriate for such usage.
Best regards,
Olivier
Rkodippily
Yes, my use case is to recognize handwritten python source code. Hard coding variable names into a custom lexicon is not really an option as you said since variable names can be anything. Thanks for the quick response!