Repository navigation
More robius-speech improvements (mostly to avoid losing word chunks) - #27
Merged
Merged
Conversation
Apple's on-device recognizer ends a segment after two seconds or so of silence, but then restarts the best transcription without setting "is final". But our swift bridge code didn't really account for that, which means it got revised into one "utterance". we now never drop any words that were already "sent" (digested?) in the segment tracker. We "commit" each chunk of words when it's over, and then make sure that we still allow shorter replacements (corrections?) to take effect and replace previously committed chunks. i tested this with a ton of different phrases and it's solid. For example, "three hundred and forty seven" does get replaced with "347" using the proper native revision.
…shown If dictation got an empty transcript after we already drew pending words, then we might end up replacing all of the pending words with that empty transcript, which looks like the drafted recognition was randomly erased. Now, across all the backends, an empty transcript doesn't change anything, only a final commit can change what was pending. Of course you can still replace the current "utterance" with revisions that come from the native service.
this makes sure we don't lose any word chunks that were already shown in the text input, even if an error happens or something else occurs that causes a final commit to never occur.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Apple's on-device recognizer ends a segment after two seconds or so
of silence, but then restarts the best transcription without
setting "is final". But our swift bridge code didn't really
account for that, which means it got revised into one "utterance".
we now never drop any words that were already "sent" (digested?)
in the segment tracker. We "commit" each chunk of words when it's over,
and then make sure that we still allow shorter replacements
(corrections?) to take effect and replace previously committed chunks.
i tested this with a ton of different phrases and it's solid.
For example, "three hundred and forty seven" does get replaced with "347"
using the proper native revision.