Title: [271958] trunk/Source/WebCore
- Revision
- 271958
- Author
- [email protected]
- Date
- 2021-01-27 09:39:23 -0800 (Wed, 27 Jan 2021)
Log Message
Use SFTranscription instead of SFTranscriptionSegment for SpeechRecognitionAlternative
https://bugs.webkit.org/show_bug.cgi?id=221026
Patch by Sihui Liu <[email protected]> on 2021-01-27
Reviewed by Youenn Fablet.
So SpeechRecognitionAlternative contains a sentence instead of a word, which matches Blink's behavior.
Manually tested in MiniBrowser.
* Modules/speech/cocoa/WebSpeechRecognizerTask.mm:
(-[WebSpeechRecognizerTaskImpl callbackWithTranscriptions:isFinal:]):
(-[WebSpeechRecognizerTaskImpl speechRecognitionTask:didHypothesizeTranscription:]):
(-[WebSpeechRecognizerTaskImpl speechRecognitionTask:didFinishRecognition:]):
(-[WebSpeechRecognizerTaskImpl callbackWithResult:isFinal:]): Deleted.
Modified Paths
Diff
Modified: trunk/Source/WebCore/ChangeLog (271957 => 271958)
--- trunk/Source/WebCore/ChangeLog 2021-01-27 17:37:36 UTC (rev 271957)
+++ trunk/Source/WebCore/ChangeLog 2021-01-27 17:39:23 UTC (rev 271958)
@@ -1,3 +1,20 @@
+2021-01-27 Sihui Liu <[email protected]>
+
+ Use SFTranscription instead of SFTranscriptionSegment for SpeechRecognitionAlternative
+ https://bugs.webkit.org/show_bug.cgi?id=221026
+
+ Reviewed by Youenn Fablet.
+
+ So SpeechRecognitionAlternative contains a sentence instead of a word, which matches Blink's behavior.
+
+ Manually tested in MiniBrowser.
+
+ * Modules/speech/cocoa/WebSpeechRecognizerTask.mm:
+ (-[WebSpeechRecognizerTaskImpl callbackWithTranscriptions:isFinal:]):
+ (-[WebSpeechRecognizerTaskImpl speechRecognitionTask:didHypothesizeTranscription:]):
+ (-[WebSpeechRecognizerTaskImpl speechRecognitionTask:didFinishRecognition:]):
+ (-[WebSpeechRecognizerTaskImpl callbackWithResult:isFinal:]): Deleted.
+
2021-01-27 Zalan Bujtas <[email protected]>
[LFC][IFC] Expand "empty line" fast path support for multiline inline boxes
Modified: trunk/Source/WebCore/Modules/speech/cocoa/WebSpeechRecognizerTask.mm (271957 => 271958)
--- trunk/Source/WebCore/Modules/speech/cocoa/WebSpeechRecognizerTask.mm 2021-01-27 17:37:36 UTC (rev 271957)
+++ trunk/Source/WebCore/Modules/speech/cocoa/WebSpeechRecognizerTask.mm 2021-01-27 17:39:23 UTC (rev 271958)
@@ -56,7 +56,7 @@
}
- (instancetype)initWithIdentifier:(SpeechRecognitionConnectionClientIdentifier)identifier locale:(NSString*)localeIdentifier doMultipleRecognitions:(BOOL)continuous reportInterimResults:(BOOL)interimResults maxAlternatives:(unsigned long)alternatives delegateCallback:(void(^)(const SpeechRecognitionUpdate&))callback;
-- (void)callbackWithResult:(SFTranscription *)transcription isFinal:(BOOL)isFinal;
+- (void)callbackWithTranscriptions:(NSArray<SFTranscription *> *)transcriptions isFinal:(BOOL)isFinal;
- (void)audioSamplesAvailable:(CMSampleBufferRef)sampleBuffer;
- (void)abort;
- (void)stop;
@@ -113,24 +113,23 @@
return self;
}
-- (void)callbackWithResult:(SFTranscription *)transcription isFinal:(BOOL)isFinal
+- (void)callbackWithTranscriptions:(NSArray<SFTranscription *> *)transcriptions isFinal:(BOOL)isFinal
{
- auto segments = [transcription segments];
- Vector<SpeechRecognitionResultData> datas;
- datas.reserveInitialCapacity(segments.count);
- for (SFTranscriptionSegment* segment in segments) {
- // Note segment confidence is reported 0 when result is not final.
- Vector<SpeechRecognitionAlternativeData> alternatives;
- alternatives.reserveInitialCapacity(_maxAlternatives);
- alternatives.uncheckedAppend(SpeechRecognitionAlternativeData { [segment substring], [segment confidence] });
- for (NSString* segmentAlternative : [segment alternativeSubstrings]) {
- if (alternatives.size() == _maxAlternatives)
- break;
- // FIXME: calculate or get alternative confidence if possible.
- alternatives.uncheckedAppend(SpeechRecognitionAlternativeData { segmentAlternative, 0.0 });
+ Vector<SpeechRecognitionAlternativeData> alternatives;
+ alternatives.reserveInitialCapacity(_maxAlternatives);
+ for (SFTranscription* transcription in transcriptions) {
+ // FIXME: <rdar://73629573> get confidence of SFTranscription when possible.
+ double maxConfidence = 0.0;
+ for (SFTranscriptionSegment* segment in [transcription segments]) {
+ double confidence = [segment confidence];
+ maxConfidence = maxConfidence < confidence ? confidence : maxConfidence;
}
- datas.uncheckedAppend(SpeechRecognitionResultData { WTFMove(alternatives), bool(isFinal) });
+ alternatives.uncheckedAppend(SpeechRecognitionAlternativeData { [transcription formattedString], maxConfidence });
+ if (alternatives.size() == _maxAlternatives)
+ break;
}
+ Vector<SpeechRecognitionResultData> datas;
+ datas.append(SpeechRecognitionResultData { WTFMove(alternatives), bool(isFinal) });
_delegateCallback(SpeechRecognitionUpdate::createResult(_identifier, WTFMove(datas)));
}
@@ -219,15 +218,14 @@
ASSERT(isMainThread());
[self sendSpeechStartIfNeeded];
- [self callbackWithResult:transcription isFinal:NO];
+ [self callbackWithTranscriptions:[NSArray arrayWithObjects:transcription, nil] isFinal:NO];
}
- (void)speechRecognitionTask:(SFSpeechRecognitionTask *)task didFinishRecognition:(SFSpeechRecognitionResult *)recognitionResult
{
ASSERT(isMainThread());
+ [self callbackWithTranscriptions:recognitionResult.transcriptions isFinal:YES];
- [self callbackWithResult:recognitionResult.bestTranscription isFinal:YES];
-
if (!_doMultipleRecognitions) {
[self sendSpeechEndIfNeeded];
[self sendEndIfNeeded];
_______________________________________________
webkit-changes mailing list
[email protected]
https://lists.webkit.org/mailman/listinfo/webkit-changes
