voice - add config for language (microsoft#205908)

* voice - add config for language

* .

* .

* .

Author: bpasero

src/vs/workbench/api/browser/mainThreadSpeech.ts

@@ -41,7 +41,7 @@ export class MainThreadSpeech implements MainThreadSpeechShape {
 
 		const registration = this.speechService.registerSpeechProvider(identifier, {
 			metadata,
-			createSpeechToTextSession: token => {
+			createSpeechToTextSession: (token, options) => {
 				if (token.isCancellationRequested) {
 					return {
 						onDidChange: Event.None
@@ -51,7 +51,7 @@ export class MainThreadSpeech implements MainThreadSpeechShape {
 				const disposables = new DisposableStore();
 				const session = Math.random();
 
-				this.proxy.$createSpeechToTextSession(handle, session);
+				this.proxy.$createSpeechToTextSession(handle, session, options?.language);
 
 				const onDidChange = disposables.add(new Emitter<ISpeechToTextEvent>());
 				this.speechToTextSessions.set(session, { onDidChange });

src/vs/workbench/api/common/extHost.protocol.ts

@@ -1171,7 +1171,7 @@ export interface MainThreadSpeechShape extends IDisposable {
 }
 
 export interface ExtHostSpeechShape {
-	$createSpeechToTextSession(handle: number, session: number): Promise<void>;
+	$createSpeechToTextSession(handle: number, session: number, language?: string): Promise<void>;
 	$cancelSpeechToTextSession(session: number): Promise<void>;
 
 	$createKeywordRecognitionSession(handle: number, session: number): Promise<void>;

src/vs/workbench/api/common/extHostSpeech.ts

@@ -24,7 +24,7 @@ export class ExtHostSpeech implements ExtHostSpeechShape {
 		this.proxy = mainContext.getProxy(MainContext.MainThreadSpeech);
 	}
 
-	async $createSpeechToTextSession(handle: number, session: number): Promise<void> {
+	async $createSpeechToTextSession(handle: number, session: number, language?: string): Promise<void> {
 		const provider = this.providers.get(handle);
 		if (!provider) {
 			return;
@@ -35,7 +35,7 @@ export class ExtHostSpeech implements ExtHostSpeechShape {
 		const cts = new CancellationTokenSource();
 		this.sessions.set(session, cts);
 
-		const speechToTextSession = disposables.add(provider.provideSpeechToTextSession(cts.token));
+		const speechToTextSession = disposables.add(provider.provideSpeechToTextSession(cts.token, language ? { language } : undefined));
 		disposables.add(speechToTextSession.onDidChange(e => {
 			if (cts.token.isCancellationRequested) {
 				return;

src/vs/workbench/contrib/accessibility/browser/accessibilityConfiguration.ts

@@ -660,9 +660,11 @@ export function registerAccessibilityConfiguration() {
 }
 
 export const enum AccessibilityVoiceSettingId {
-	SpeechTimeout = 'accessibility.voice.speechTimeout'
+	SpeechTimeout = 'accessibility.voice.speechTimeout',
+	SpeechLanguage = 'accessibility.voice.speechLanguage'
 }
 export const SpeechTimeoutDefault = 1200;
+const SpeechLanguageDefault = 'en-US';
 
 export class DynamicSpeechAccessibilityConfiguration extends Disposable implements IWorkbenchContribution {
 
@@ -681,6 +683,11 @@ export class DynamicSpeechAccessibilityConfiguration extends Disposable implemen
 			return; // these settings require a speech provider
 		}
 
+		const languages = this.getLanguages();
+		const languagesSorted = Object.keys(languages).sort((langA, langB) => {
+			return languages[langA].name.localeCompare(languages[langB].name);
+		});
+
 		const registry = Registry.as<IConfigurationRegistry>(Extensions.Configuration);
 		registry.registerConfiguration({
 			...accessibilityConfigurationNodeBase,
@@ -691,11 +698,98 @@ export class DynamicSpeechAccessibilityConfiguration extends Disposable implemen
 					'default': SpeechTimeoutDefault,
 					'minimum': 0,
 					'tags': ['accessibility']
+				},
+				[AccessibilityVoiceSettingId.SpeechLanguage]: {
+					'markdownDescription': localize('voice.speechLanguage', "The language that voice speech recognition should recognize."),
+					'type': 'string',
+					'enum': languagesSorted,
+					'default': SpeechLanguageDefault,
+					'tags': ['accessibility'],
+					'enumDescriptions': languagesSorted.map(key => languages[key].name),
+					'enumItemLabels': languagesSorted.map(key => languages[key].name)
 				}
 			}
 		});
 	}
+
+	private getLanguages(): { [locale: string]: { name: string } } {
+		return {
+			['de-DE']: {
+				name: localize('speechLanguage.de-DE', "German (Germany)")
+			},
+			['en-AU']: {
+				name: localize('speechLanguage.en-AU', "English (Australia)")
+			},
+			['en-CA']: {
+				name: localize('speechLanguage.en-CA', "English (Canada)")
+			},
+			['en-GB']: {
+				name: localize('speechLanguage.en-GB', "English (United Kingdom)")
+			},
+			['en-IE']: {
+				name: localize('speechLanguage.en-IE', "English (Ireland)")
+			},
+			['en-IN']: {
+				name: localize('speechLanguage.en-IN', "English (India)")
+			},
+			['en-NZ']: {
+				name: localize('speechLanguage.en-NZ', "English (New Zealand)")
+			},
+			[SpeechLanguageDefault]: {
+				name: localize('speechLanguage.en-US', "English (United States)")
+			},
+			['es-ES']: {
+				name: localize('speechLanguage.es-ES', "Spanish (Spain)")
+			},
+			['es-MX']: {
+				name: localize('speechLanguage.es-MX', "Spanish (Mexico)")
+			},
+			['fr-CA']: {
+				name: localize('speechLanguage.fr-CA', "French (Canada)")
+			},
+			['fr-FR']: {
+				name: localize('speechLanguage.fr-FR', "French (France)")
+			},
+			['hi-IN']: {
+				name: localize('speechLanguage.hi-IN', "Hindi (India)")
+			},
+			['it-IT']: {
+				name: localize('speechLanguage.it-IT', "Italian (Italy)")
+			},
+			['ja-JP']: {
+				name: localize('speechLanguage.ja-JP', "Japanese (Japan)")
+			},
+			['ko-KR']: {
+				name: localize('speechLanguage.ko-KR', "Korean (South Korea)")
+			},
+			['nl-NL']: {
+				name: localize('speechLanguage.nl-NL', "Dutch (Netherlands)")
+			},
+			['pt-BR']: {
+				name: localize('speechLanguage.pt-BR', "Portuguese (Brazil)")
+			},
+			['ru-RU']: {
+				name: localize('speechLanguage.ru-RU', "Russian (Russia)")
+			},
+			['sv-SE']: {
+				name: localize('speechLanguage.sv-SE', "Swedish (Sweden)")
+			},
+			['tr-TR']: {
+				name: localize('speechLanguage.tr-TR', "Turkish (Turkey)")
+			},
+			['zh-CN']: {
+				name: localize('speechLanguage.zh-CN', "Chinese (Simplified, China)")
+			},
+			['zh-HK']: {
+				name: localize('speechLanguage.zh-HK', "Chinese (Traditional, Hong Kong)")
+			},
+			['zh-TW']: {
+				name: localize('speechLanguage.zh-TW', "Chinese (Traditional, Taiwan)")
+			}
+		};
+	}
 }
+
 Registry.as<IConfigurationMigrationRegistry>(WorkbenchExtensions.ConfigurationMigration)
 	.registerConfigurationMigrations([{
 		key: 'audioCues.volume',

src/vs/workbench/contrib/speech/browser/speechService.ts

@@ -13,6 +13,7 @@ import { IHostService } from 'vs/workbench/services/host/browser/host';
 import { DeferredPromise } from 'vs/base/common/async';
 import { ISpeechService, ISpeechProvider, HasSpeechProvider, ISpeechToTextSession, SpeechToTextInProgress, IKeywordRecognitionSession, KeywordRecognitionStatus, SpeechToTextStatus } from 'vs/workbench/contrib/speech/common/speechService';
 import { ITelemetryService } from 'vs/platform/telemetry/common/telemetry';
+import { IConfigurationService } from 'vs/platform/configuration/common/configuration';
 
 export class SpeechService extends Disposable implements ISpeechService {
 
@@ -34,7 +35,8 @@ export class SpeechService extends Disposable implements ISpeechService {
 		@ILogService private readonly logService: ILogService,
 		@IContextKeyService private readonly contextKeyService: IContextKeyService,
 		@IHostService private readonly hostService: IHostService,
-		@ITelemetryService private readonly telemetryService: ITelemetryService
+		@ITelemetryService private readonly telemetryService: ITelemetryService,
+		@IConfigurationService private readonly configurationService: IConfigurationService
 	) {
 		super();
 	}
@@ -78,7 +80,8 @@ export class SpeechService extends Disposable implements ISpeechService {
 			this.logService.warn(`Multiple speech providers registered. Picking first one: ${provider.metadata.displayName}`);
 		}
 
-		const session = this._activeSpeechToTextSession = provider.createSpeechToTextSession(token);
+		const language = this.configurationService.getValue<string>('accessibility.voice.speechLanguage');
+		const session = this._activeSpeechToTextSession = provider.createSpeechToTextSession(token, typeof language === 'string' ? { language } : undefined);
 
 		const sessionStart = Date.now();
 		let sessionRecognized = false;

src/vs/workbench/contrib/speech/common/speechService.ts

@@ -52,10 +52,14 @@ export interface IKeywordRecognitionSession {
 	readonly onDidChange: Event<IKeywordRecognitionEvent>;
 }
 
+export interface ISpeechToTextSessionOptions {
+	readonly language?: string;
+}
+
 export interface ISpeechProvider {
 	readonly metadata: ISpeechProviderMetadata;
 
-	createSpeechToTextSession(token: CancellationToken): ISpeechToTextSession;
+	createSpeechToTextSession(token: CancellationToken, options?: ISpeechToTextSessionOptions): ISpeechToTextSession;
 	createKeywordRecognitionSession(token: CancellationToken): IKeywordRecognitionSession;
 }
 

src/vscode-dts/vscode.proposed.speech.d.ts

@@ -7,6 +7,10 @@ declare module 'vscode' {
 
 	// todo@bpasero work in progress speech API
 
+	export interface SpeechToTextOptions {
+		readonly language?: string;
+	}
+
 	export enum SpeechToTextStatus {
 		Started = 1,
 		Recognizing = 2,
@@ -38,7 +42,7 @@ declare module 'vscode' {
 	}
 
 	export interface SpeechProvider {
-		provideSpeechToTextSession(token: CancellationToken): SpeechToTextSession;
+		provideSpeechToTextSession(token: CancellationToken, options?: SpeechToTextOptions): SpeechToTextSession;
 		provideKeywordRecognitionSession(token: CancellationToken): KeywordRecognitionSession;
 	}