ML.NETで誰が言ったセリフかを推定する

whoistalking.cs

// ML.NETで誰が言ったセリフかを推定する
//
// ML.NETのセットアップ方法は下記を参照
// https://www.microsoft.com/net/learn/apps/machine-learning-and-ai/ml-dotnet/get-started/windows
using Microsoft.ML;
using Microsoft.ML.Models;
using Microsoft.ML.Runtime.Api;
using Microsoft.ML.Trainers;
using Microsoft.ML.Transforms;
using System;

namespace whoistalking
{
    class Script
    {
        [Column(ordinal: "0")]
        public string Character;
        [Column(ordinal: "1")]
        public string Line;
    }
    public class ScriptPrediction
    {
        [ColumnName("PredictedLabel")]
        public string Character;
    }
    class Program
    {
        static void Main(string[] args)
        {
            // 話者名\t台詞 を1行としたtsvファイルを用意する
            var data = new TextLoader<Script>(@"zero.tsv.data.tsv", useHeader: false, separator: "tab");
            var testData = new TextLoader<Script>(@"zero.tsv.test.tsv", useHeader: false, separator: "tab");

            // LearningPipelineの作成
            var pipeline = new LearningPipeline();
            pipeline.Add(data);
            pipeline.Add(new Dictionarizer(("Character", "Label")));
            pipeline.Add(new TextFeaturizer("Line", "Line"));
            pipeline.Add(new ColumnConcatenator("Features", "Line"));
            pipeline.Add(new StochasticDualCoordinateAscentClassifier());
            pipeline.Add(new PredictedLabelColumnOriginalValueConverter() { PredictedLabelColumn = "PredictedLabel" });

            // 学習
            PredictionModel<Script, ScriptPrediction> model = pipeline.Train<Script, ScriptPrediction>();

            // テスト用データによる評価
            var evaluator = new ClassificationEvaluator();
            ClassificationMetrics metrics = evaluator.Evaluate(model, testData);
            Console.WriteLine("Micro-Accuracy: {0}", metrics.AccuracyMicro);

            // キーボードからの入力
            while (true)
            {
                Console.Write("台詞: ");
                string test = Console.ReadLine();
                Script issue = new Script
                {
                    Character = "",
                    Line = test
                };
                ScriptPrediction prediction = model.Predict(issue);
                Console.WriteLine("推定話者: " + prediction.Character);
            }
        }
    }

}