I'm currently working on an android application targeting young people for educational purposes.
In this application I would like to be able to retrieve what the child is writing on the tablet without your engine replacing the input by the closest word it thinks the child has written.
For example, when I type solation, i would like that the engine tells me that I typed solation and not solution.
The purpose of the app being to teach the child how to write correctly, I need to detect his mistakes.
Do you know if what I'm describing is possible with your engine ? (So far, I have tested the ATK samples and failed to see an option to disable this behavior) If so, can you give me hints on how to achieve this ?
Actually as mentioned in another post, the easiest way to allow 'out of lexicon' results would be by removing the linguistic knowledge model of our Text widgets, but it will also impact global recognition accuracy as our recognition engine has been trained thanks to this lk file.
We also offer now a tool for registered users to create custom resources that may interest you:
Thanks for your answer, that's exactly what I needed !