-
Notifications
You must be signed in to change notification settings - Fork 349
Handle low confidence spacing and spelling #249
Copy link
Copy link
Open
Description
Great project, thanks for your work.
I have already seen some improvement reducing the window size to 1000, and in the chunk files generated by CORRECT_TEXT most of the confidence levels are 1.0 but some still have confidence levels between 0.90 - 0.99, where words are missing spaces between them or have misspellings.
What would be the best way to automatically handle this?
After correct text runs on a chunk, if the confidence level is less than 1.0 should I make it reiterate reducing the window size 1000 - 800 - 600 - 400 - 200, or is there a better way?
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
No labels