speech service demo: STT

Program.cs

using System;
using System.Threading.Tasks;
using Microsoft.CognitiveServices.Speech;

namespace SpeechServiceConsole
{
    class Program
    {
        private const string SubscriptionKey = "????????????"; // TODO: input !!
        private const string Region = "??????????????????"; // TODO: input !!

        private const ConsoleColor DefaultColor = ConsoleColor.White;
        private const ConsoleColor ProcessingColor = ConsoleColor.Yellow;
        private const ConsoleColor AppMessageColor = ConsoleColor.Green;

        static async Task Main(string[] args)
        {
            ConsoleKey inputKey;

            do
            {
                ShowMenu();
                inputKey = Console.ReadKey().Key;


                switch (inputKey)
                {
                    case ConsoleKey.D1:
                        await RunContinuousSpeechRecognitionAsync();
                        break;

                    case ConsoleKey.D2:
                        await RunSpeechTranslationAsync();
                        break;
                    default:
                        Console.WriteLine();
                        Console.WriteLine("入力不正です。メニューから数字を選んでください。");
                        ShowMenu();
                        break;
                }

            } while (inputKey != ConsoleKey.D0);
        }



        private static void ShowMenu()
        {
            Console.WriteLine();
            Console.WriteLine("メニューの数字を選んでください。");

            Console.WriteLine("1. ダラダラと会話をテキスト化する");
            Console.WriteLine("2. ダラダラと会話を翻訳する");

            Console.WriteLine("0. 終了");
        }

        private static readonly TaskCompletionSource<int> StopRecognition = new TaskCompletionSource<int>();


        private static async Task RunContinuousSpeechRecognitionAsync()
        {
            var language = SelectLanguage("話す言語を数字で選んでください");

            var config = SpeechConfig.FromSubscription(SubscriptionKey, Region);
            config.SpeechRecognitionLanguage = language;

            var recognizer = new SpeechRecognizer(config);

            SetEventHandler(recognizer);

            await recognizer.StartContinuousRecognitionAsync().ConfigureAwait(false);

            Task.WaitAny(StopRecognition.Task);
            await recognizer.StopContinuousRecognitionAsync().ConfigureAwait(false);
        }


        private static async Task RunSpeechTranslationAsync()
        {
            // 未実装ですよー
            throw new NotImplementedException();
        }

        #region Language Selection

        private static string SelectLanguage(string message)
        {
            Console.WriteLine();
            Console.WriteLine(message);
            ShowLanguageMenu();
            Console.WriteLine();

            while (true)
            {
                var input = Console.ReadKey().Key;

                // 参考: 言語コードリスト https://docs.microsoft.com/ja-jp/azure/cognitive-services/speech-service/language-support
                switch (input)
                {
                    case ConsoleKey.D1: return "ja-jp";
                    case ConsoleKey.D2: return "en-us";

                    default:
                        Console.ForegroundColor = DefaultColor;
                        Console.WriteLine();
                        Console.WriteLine("Oh No! メニューから数字を選んでください。");
                        ShowLanguageMenu();
                        break;
                }
            }
        }


        private static void ShowLanguageMenu()
        {

            Console.WriteLine("1. 日本語");
            Console.WriteLine("2. English");
        }

        #endregion


        private static void SetEventHandler(SpeechRecognizer recognizer)
        {
            recognizer.SessionStarted += (s, e) =>
            {
                Console.ForegroundColor = AppMessageColor;
                Console.WriteLine();
                Console.WriteLine("SESSION-STARTED.: 独り言をマイクに向かって呟いてみましょう！「ストップ」というと停止します。");
                Console.WriteLine("----------------------------------");
            };

            recognizer.Recognizing += (s, e) =>
            {
                Console.ForegroundColor = ProcessingColor;
                Console.WriteLine($"RECOGNIZING: {e.Result.Text}");
            };

            recognizer.Recognized += (s, e) =>
            {
                if (e.Result.Reason == ResultReason.RecognizedSpeech)
                {
                    Console.ForegroundColor = DefaultColor;
                    Console.WriteLine($"RECOGNIZED: {e.Result.Text}");

                    if (IsStopWord(e.Result.Text))
                    {
                        StopRecognition.TrySetResult(0);
                    }
                }
                else
                {
                    Console.ForegroundColor = AppMessageColor;
                    Console.WriteLine($"NOMATCH: 認識できませんでした。");
                }

                Console.WriteLine();
            };

            recognizer.Canceled += (s, e) =>
            {
                Console.ForegroundColor = AppMessageColor;
                Console.WriteLine($"CANCELED: Reason={e.Reason}");
                Console.WriteLine();

                if (e.Reason == CancellationReason.Error)
                {
                    Console.WriteLine($"CANCELED: ErrorCode={e.ErrorCode}; ErrorDetails={e.ErrorDetails};");
                }
                StopRecognition.TrySetResult(0);
            };

            recognizer.SessionStopped += (s, e) =>
            {
                StopRecognition.TrySetResult(0);
                Console.WriteLine("SESSION-STOPPED: executed SessionStopped event handler.");
            };
        }


        private static bool IsStopWord(string message)
        {
            return message == "ストップ。" || message == "Stop.";
        }

    }
}