[Feature] Replace Spoken Language Identification with MediaPipe Solutions #116
Labels
enhancement
New feature or request
good first issue
Good for newcomers
help wanted
Extra attention is needed
spoken-to-signed
Problem
The current spoken language identification - implemented in #10 - has limitations, specifically, it struggles with short texts
google/cld3#76 (comment)
Description
The MediaPipe Language identification does not suffer from the same limitations. It is also smaller in size (315kb compared to 1.1mb).
We can replace the current spoken language identification, or better yet, allow for options in the settings.
Alternatives
Keep the current identification. For spoken language change recommendation, don't show the suggestion until the text is "long enough"
![image](https://private-user-images.githubusercontent.com/5757359/281045463-4b4c7c35-282a-4bb5-92e4-456e4ccf5e28.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.fiTiCRXPmn3CLnF1BR5wVdM6H3obxrkHbtHBVzFlWYQ)
The text was updated successfully, but these errors were encountered: