UNPKG

microsoft-cognitiveservices-speech-sdk

Version:
1 lines 11.2 kB
{"version":3,"sources":["src/common.speech/SpeechServiceRecognizer.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,YAAY,EAAE,MAAM,sBAAsB,CAAC;AACpD,OAAO,EACH,qBAAqB,EACrB,kBAAkB,EAQlB,gBAAgB,EACnB,MAAM,mBAAmB,CAAC;AAC3B,OAAO,EAMH,qBAAqB,EAGxB,MAAM,cAAc,CAAC;AACtB,OAAO,EAAE,eAAe,EAAE,MAAM,sBAAsB,CAAC;AACvD,OAAO,EAAE,kBAAkB,EAAE,MAAM,yBAAyB,CAAC;AAC7D,OAAO,EAAE,gBAAgB,EAAE,MAAM,uBAAuB,CAAC;AACzD,OAAO,EAAE,uBAAuB,EAAE,MAAM,uCAAuC,CAAC;AAGhF,qBAAa,uBAAwB,SAAQ,qBAAqB;IAE9D,OAAO,CAAC,oBAAoB,CAAmB;gBAG3C,cAAc,EAAE,eAAe,EAC/B,iBAAiB,EAAE,kBAAkB,EACrC,WAAW,EAAE,YAAY,EACzB,gBAAgB,EAAE,gBAAgB,EAClC,gBAAgB,EAAE,gBAAgB;cAMtB,2BAA2B,CAAC,iBAAiB,EAAE,uBAAuB,GAAG,OAAO,CAAC,OAAO,CAAC;IAqIzG,SAAS,CAAC,iBAAiB,CACvB,SAAS,EAAE,MAAM,EACjB,SAAS,EAAE,MAAM,EACjB,kBAAkB,EAAE,kBAAkB,EACtC,SAAS,EAAE,qBAAqB,EAChC,KAAK,EAAE,MAAM,GAAG,IAAI;CAsC3B","file":"SpeechServiceRecognizer.d.ts","sourcesContent":["// Copyright (c) Microsoft Corporation. All rights reserved.\r\n// Licensed under the MIT license.\r\n\r\nimport { IAudioSource } from \"../common/Exports.js\";\r\nimport {\r\n CancellationErrorCode,\r\n CancellationReason,\r\n OutputFormat,\r\n PropertyCollection,\r\n PropertyId,\r\n ResultReason,\r\n SpeechRecognitionCanceledEventArgs,\r\n SpeechRecognitionEventArgs,\r\n SpeechRecognitionResult,\r\n SpeechRecognizer,\r\n} from \"../sdk/Exports.js\";\r\nimport {\r\n CancellationErrorCodePropertyName,\r\n DetailedSpeechPhrase,\r\n EnumTranslation,\r\n OutputFormatPropertyName,\r\n RecognitionStatus,\r\n ServiceRecognizerBase,\r\n SimpleSpeechPhrase,\r\n SpeechHypothesis,\r\n} from \"./Exports.js\";\r\nimport { IAuthentication } from \"./IAuthentication.js\";\r\nimport { IConnectionFactory } from \"./IConnectionFactory.js\";\r\nimport { RecognizerConfig } from \"./RecognizerConfig.js\";\r\nimport { SpeechConnectionMessage } from \"./SpeechConnectionMessage.Internal.js\";\r\n\r\n// eslint-disable-next-line max-classes-per-file\r\nexport class SpeechServiceRecognizer extends ServiceRecognizerBase {\r\n\r\n private privSpeechRecognizer: SpeechRecognizer;\r\n\r\n public constructor(\r\n authentication: IAuthentication,\r\n connectionFactory: IConnectionFactory,\r\n audioSource: IAudioSource,\r\n recognizerConfig: RecognizerConfig,\r\n speechRecognizer: SpeechRecognizer) {\r\n super(authentication, connectionFactory, audioSource, recognizerConfig, speechRecognizer);\r\n this.privSpeechRecognizer = speechRecognizer;\r\n\r\n }\r\n\r\n protected async processTypeSpecificMessages(connectionMessage: SpeechConnectionMessage): Promise<boolean> {\r\n\r\n let result: SpeechRecognitionResult;\r\n\r\n const resultProps: PropertyCollection = new PropertyCollection();\r\n\r\n let processed: boolean = false;\r\n\r\n switch (connectionMessage.path.toLowerCase()) {\r\n case \"speech.hypothesis\":\r\n case \"speech.fragment\":\r\n const hypothesis: SpeechHypothesis = SpeechHypothesis.fromJSON(connectionMessage.textBody, this.privRequestSession.currentTurnAudioOffset);\r\n resultProps.setProperty(PropertyId.SpeechServiceResponse_JsonResult, hypothesis.asJson());\r\n\r\n result = new SpeechRecognitionResult(\r\n this.privRequestSession.requestId,\r\n ResultReason.RecognizingSpeech,\r\n hypothesis.Text,\r\n hypothesis.Duration,\r\n hypothesis.Offset,\r\n hypothesis.Language,\r\n hypothesis.LanguageDetectionConfidence,\r\n undefined, // Speaker Id\r\n undefined,\r\n hypothesis.asJson(),\r\n resultProps);\r\n\r\n this.privRequestSession.onHypothesis(hypothesis.Offset);\r\n\r\n const ev = new SpeechRecognitionEventArgs(result, hypothesis.Offset, this.privRequestSession.sessionId);\r\n\r\n if (!!this.privSpeechRecognizer.recognizing) {\r\n try {\r\n this.privSpeechRecognizer.recognizing(this.privSpeechRecognizer, ev);\r\n /* eslint-disable no-empty */\r\n } catch (error) {\r\n // Not going to let errors in the event handler\r\n // trip things up.\r\n }\r\n }\r\n processed = true;\r\n break;\r\n case \"speech.phrase\":\r\n const simple: SimpleSpeechPhrase = SimpleSpeechPhrase.fromJSON(connectionMessage.textBody, this.privRequestSession.currentTurnAudioOffset);\r\n resultProps.setProperty(PropertyId.SpeechServiceResponse_JsonResult, simple.asJson());\r\n\r\n const resultReason: ResultReason = EnumTranslation.implTranslateRecognitionResult(simple.RecognitionStatus, this.privExpectContentAssessmentResponse);\r\n\r\n this.privRequestSession.onPhraseRecognized(simple.Offset + simple.Duration);\r\n\r\n if (ResultReason.Canceled === resultReason) {\r\n const cancelReason: CancellationReason = EnumTranslation.implTranslateCancelResult(simple.RecognitionStatus);\r\n const cancellationErrorCode: CancellationErrorCode = EnumTranslation.implTranslateCancelErrorCode(simple.RecognitionStatus);\r\n\r\n await this.cancelRecognitionLocal(\r\n cancelReason,\r\n cancellationErrorCode,\r\n EnumTranslation.implTranslateErrorDetails(cancellationErrorCode));\r\n\r\n } else {\r\n // Like the native SDK's, don't event / return an EndOfDictation message.\r\n if (simple.RecognitionStatus === RecognitionStatus.EndOfDictation) {\r\n break;\r\n }\r\n\r\n if (this.privRecognizerConfig.parameters.getProperty(OutputFormatPropertyName) === OutputFormat[OutputFormat.Simple]) {\r\n result = new SpeechRecognitionResult(\r\n this.privRequestSession.requestId,\r\n resultReason,\r\n simple.DisplayText,\r\n simple.Duration,\r\n simple.Offset,\r\n simple.Language,\r\n simple.LanguageDetectionConfidence,\r\n undefined, // Speaker Id\r\n undefined,\r\n simple.asJson(),\r\n resultProps);\r\n } else {\r\n const detailed: DetailedSpeechPhrase = DetailedSpeechPhrase.fromJSON(connectionMessage.textBody, this.privRequestSession.currentTurnAudioOffset);\r\n resultProps.setProperty(PropertyId.SpeechServiceResponse_JsonResult, detailed.asJson());\r\n\r\n result = new SpeechRecognitionResult(\r\n this.privRequestSession.requestId,\r\n resultReason,\r\n detailed.RecognitionStatus === RecognitionStatus.Success ? detailed.NBest[0].Display : \"\",\r\n detailed.Duration,\r\n detailed.Offset,\r\n detailed.Language,\r\n detailed.LanguageDetectionConfidence,\r\n undefined, // Speaker Id\r\n undefined,\r\n detailed.asJson(),\r\n resultProps);\r\n }\r\n\r\n const event: SpeechRecognitionEventArgs = new SpeechRecognitionEventArgs(result, result.offset, this.privRequestSession.sessionId);\r\n\r\n if (!!this.privSpeechRecognizer.recognized) {\r\n try {\r\n this.privSpeechRecognizer.recognized(this.privSpeechRecognizer, event);\r\n /* eslint-disable no-empty */\r\n } catch (error) {\r\n // Not going to let errors in the event handler\r\n // trip things up.\r\n }\r\n }\r\n\r\n\r\n if (!!this.privSuccessCallback) {\r\n try {\r\n this.privSuccessCallback(result);\r\n } catch (e) {\r\n if (!!this.privErrorCallback) {\r\n this.privErrorCallback(e as string);\r\n }\r\n }\r\n // Only invoke the call back once.\r\n // and if it's successful don't invoke the\r\n // error after that.\r\n this.privSuccessCallback = undefined;\r\n this.privErrorCallback = undefined;\r\n }\r\n }\r\n processed = true;\r\n break;\r\n default:\r\n break;\r\n }\r\n return processed;\r\n }\r\n\r\n // Cancels recognition.\r\n protected cancelRecognition(\r\n sessionId: string,\r\n requestId: string,\r\n cancellationReason: CancellationReason,\r\n errorCode: CancellationErrorCode,\r\n error: string): void {\r\n\r\n const properties: PropertyCollection = new PropertyCollection();\r\n properties.setProperty(CancellationErrorCodePropertyName, CancellationErrorCode[errorCode]);\r\n\r\n if (!!this.privSpeechRecognizer.canceled) {\r\n const cancelEvent: SpeechRecognitionCanceledEventArgs = new SpeechRecognitionCanceledEventArgs(\r\n cancellationReason,\r\n error,\r\n errorCode,\r\n undefined,\r\n sessionId);\r\n try {\r\n this.privSpeechRecognizer.canceled(this.privSpeechRecognizer, cancelEvent);\r\n /* eslint-disable no-empty */\r\n } catch { }\r\n }\r\n\r\n if (!!this.privSuccessCallback) {\r\n const result: SpeechRecognitionResult = new SpeechRecognitionResult(\r\n requestId,\r\n ResultReason.Canceled,\r\n undefined, // Text\r\n undefined, // Duration\r\n undefined, // Offset\r\n undefined, // Language\r\n undefined, // Language Detection Confidence\r\n undefined, // Speaker Id\r\n error,\r\n undefined, // Json\r\n properties);\r\n try {\r\n this.privSuccessCallback(result);\r\n this.privSuccessCallback = undefined;\r\n /* eslint-disable no-empty */\r\n } catch { }\r\n }\r\n }\r\n}\r\n"]}