change

src/common.speech/ConversationTranscriptionServiceRecognizer.ts

@@ -28,6 +28,7 @@ import { IAuthentication } from "./IAuthentication";
 import { IConnectionFactory } from "./IConnectionFactory";
 import { RecognizerConfig } from "./RecognizerConfig";
 import { SpeechConnectionMessage } from "./SpeechConnectionMessage.Internal";
+import { PhraseDetection, SpeakerDiarization } from "./ServiceRecognizerBase";
 
 // eslint-disable-next-line max-classes-per-file
 export class ConversationTranscriptionServiceRecognizer extends ServiceRecognizerBase {
@@ -40,8 +41,21 @@ export class ConversationTranscriptionServiceRecognizer extends ServiceRecognize
         audioSource: IAudioSource,
         recognizerConfig: RecognizerConfig,
         conversationTranscriber: ConversationTranscriber) {
-        super(authentication, connectionFactory, audioSource, recognizerConfig, conversationTranscriber, true);
+        super(authentication, connectionFactory, audioSource, recognizerConfig, conversationTranscriber);
         this.privConversationTranscriber = conversationTranscriber;
+        this.setSpeakerDiarizationJson();
+    }
+
+    protected setSpeakerDiarizationJson(): void {
+        if (this.privEnableSpeakerId) {
+            const phraseDetection = this.privSpeechContext.getSection("phraseDetection") as PhraseDetection;
+            const speakerDiarization: SpeakerDiarization = {};
+            speakerDiarization.mode = "Anonymous";
+            speakerDiarization.audioSessionId = this.privDiarizationSessionId;
+            speakerDiarization.audioOffsetMs = 0;
+            phraseDetection.speakerDiarization = speakerDiarization;
+            this.privSpeechContext.setSection("phraseDetection", phraseDetection);
+        }
     }
 
     protected async processTypeSpecificMessages(connectionMessage: SpeechConnectionMessage): Promise<boolean> {

src/common.speech/RecognizerConfig.ts

@@ -23,6 +23,7 @@ export class RecognizerConfig {
     private privRecognitionActivityTimeout: number;
     private privParameters: PropertyCollection;
     private privMaxRetryCount: number;
+    private privEnableSpeakerId: boolean;
 
     public constructor(
         speechServiceConfig: SpeechServiceConfig,
@@ -31,6 +32,7 @@ export class RecognizerConfig {
         this.privParameters = parameters;
         this.privMaxRetryCount = parseInt(parameters.getProperty("SPEECH-Error-MaxRetryCount", "4"), 10);
         this.privLanguageIdMode = parameters.getProperty(PropertyId.SpeechServiceConnection_LanguageIdMode, undefined);
+        this.privEnableSpeakerId = false;
     }
 
     public get parameters(): PropertyCollection {
@@ -92,6 +94,14 @@ export class RecognizerConfig {
     public get maxRetryCount(): number {
         return this.privMaxRetryCount;
     }
+
+    public get isSpeakerDiarizationEnabled(): boolean {
+        return this.privEnableSpeakerId;
+    }
+
+    public set isSpeakerDiarizationEnabled(value: boolean) {
+        this.privEnableSpeakerId = value;
+    }
 }
 
 // The config is serialized and sent as the Speech.Config

src/common.speech/ServiceRecognizerBase.ts

@@ -57,7 +57,7 @@ interface CustomModel {
     endpoint: string;
 }
 
-interface PhraseDetection {
+export interface PhraseDetection {
     customModels?: CustomModel[];
     onInterim?: { action: string };
     onSuccess?: { action: string };
@@ -68,14 +68,14 @@ interface PhraseDetection {
     speakerDiarization?: SpeakerDiarization;
 }
 
-interface SpeakerDiarization {
+export interface SpeakerDiarization {
     mode?: string;
     audioSessionId?: string;
     audioOffsetMs?: number;
     identityProvider?: string;
 }
 
-interface Segmentation {
+export interface Segmentation {
     segmentation: {
         mode: "Custom";
         segmentationSilenceTimeoutMs: number;
@@ -121,8 +121,7 @@ export abstract class ServiceRecognizerBase implements IDisposable {
         connectionFactory: IConnectionFactory,
         audioSource: IAudioSource,
         recognizerConfig: RecognizerConfig,
-        recognizer: Recognizer,
-        enableSpeakerId: boolean = false) {
+        recognizer: Recognizer) {
 
         if (!authentication) {
             throw new ArgumentNullError("authentication");
@@ -140,7 +139,7 @@ export abstract class ServiceRecognizerBase implements IDisposable {
             throw new ArgumentNullError("recognizerConfig");
         }
 
-        this.privEnableSpeakerId = enableSpeakerId;
+        this.privEnableSpeakerId = recognizerConfig.isSpeakerDiarizationEnabled;
         this.privMustReportEndOfStream = false;
         this.privAuthentication = authentication;
         this.privConnectionFactory = connectionFactory;
@@ -176,12 +175,10 @@ export abstract class ServiceRecognizerBase implements IDisposable {
 
         if (this.privEnableSpeakerId) {
             this.privDiarizationSessionId = createNoDashGuid();
-            // this.setSpeakerDiarizationJson(phraseDetection);
         }
 
         this.setLanguageIdJson();
         this.setOutputDetailLevelJson();
-        this.setSpeakerDiarizationJson();
     }
 
     protected setSpeechSegmentationTimeoutJson(): void{
@@ -202,20 +199,6 @@ export abstract class ServiceRecognizerBase implements IDisposable {
         }
     }
 
-    protected setSpeakerDiarizationJson(): void {
-        if (this.privEnableSpeakerId) {
-            const phraseDetection = this.privSpeechContext.getSection("phraseDetection") as PhraseDetection;
-            const speakerDiarization: SpeakerDiarization = {};
-            speakerDiarization.mode = "Anonymous";
-            speakerDiarization.audioSessionId = this.privDiarizationSessionId;
-            // eslint-disable-next-line no-console
-            console.log("***** ServiceRecognizerBase sessionId: " + this.privDiarizationSessionId);
-            speakerDiarization.audioOffsetMs = 0;
-            phraseDetection.speakerDiarization = speakerDiarization;
-            this.privSpeechContext.setSection("phraseDetection", phraseDetection);
-        }
-    }
-
     protected setLanguageIdJson(): void {
         if (this.privRecognizerConfig.autoDetectSourceLanguages !== undefined) {
             const phraseDetection = this.privSpeechContext.getSection("phraseDetection") as PhraseDetection;

src/sdk/Transcription/ConversationTranscriber.ts

@@ -235,6 +235,7 @@ export class ConversationTranscriber extends Recognizer {
         audioConfig: AudioConfig,
         recognizerConfig: RecognizerConfig): ServiceRecognizerBase {
         const configImpl: AudioConfigImpl = audioConfig as AudioConfigImpl;
+        recognizerConfig.isSpeakerDiarizationEnabled = true;
         return new ConversationTranscriptionServiceRecognizer(authentication, connectionFactory, configImpl, recognizerConfig, this);
     }
 }