Products

Data Converter Cleaner TUC - Unit Converter Import Export Survey Manager Tweet Controller

Developer

Article Blog Forum Interview Beginner News Startup Quiz

Users

Career

Apply Job Internship

Pricing

Login Sign Up Get a Quote

How can you perform sentiment analysis using C# and ML.NET?

Anubhav Sharma 1432 06 Jul 2026

How can you perform sentiment analysis using C# and ML.NET?

Ask by Anubhav Sharma

Updated 17 Aug 2026

1 Answers

Yogendra Mohan 17 Aug 2026

You can perform sentiment analysis in C# with ML.NET by training a binary classification model that predicts whether a piece of text is positive or negative.

1. Install ML.NET

Create a console project and add the ML.NET package:

dotnet new console -n SentimentAnalysis
cd SentimentAnalysis
dotnet add package Microsoft.ML

2. Create the data model

Suppose your training data is a CSV file containing:

Sentiment,Text
true,I absolutely love this product!
false,This was a terrible experience.
true,The service was excellent.
false,I would not recommend this.

Define C# classes:

using Microsoft.ML.Data;

public class SentimentData
{
    [LoadColumn(0)]
    public bool Sentiment { get; set; }

    [LoadColumn(1)]
    public string Text { get; set; }
}

public class SentimentPrediction
{
    [ColumnName("PredictedLabel")]
    public bool Prediction { get; set; }

    public float Probability { get; set; }

    public float Score { get; set; }
}

3. Load and split the data

using Microsoft.ML;

MLContext mlContext = new MLContext();

IDataView data = mlContext.Data.LoadFromTextFile<SentimentData>(
    "sentiment-data.csv",
    hasHeader: true,
    separatorChar: ',');

DataOperationsCatalog.TrainTestData split =
    mlContext.Data.TrainTestSplit(data, testFraction: 0.2);

This keeps some data aside for evaluating the model after training.

4. Build the ML.NET pipeline

For text sentiment analysis, you can convert the text into numerical features and then train a binary classifier:

IEstimator<ITransformer> pipeline =
    mlContext.Transforms.Text.FeaturizeText(
        outputColumnName: "Features",
        inputColumnName: nameof(SentimentData.Text))
    .Append(
        mlContext.BinaryClassification.Trainers.SdcaLogisticRegression(
            labelColumnName: nameof(SentimentData.Sentiment),
            featureColumnName: "Features"));

FeaturizeText performs the text-processing/feature-extraction step, while logistic regression learns how those features relate to positive or negative sentiment.

5. Train the model

ITransformer model = pipeline.Fit(split.TrainSet);

6. Evaluate it

IDataView predictions = model.Transform(split.TestSet);

CalibratedBinaryClassificationMetrics metrics =
    mlContext.BinaryClassification.Evaluate(
        predictions,
        labelColumnName: nameof(SentimentData.Sentiment));

Console.WriteLine($"Accuracy: {metrics.Accuracy:P2}");
Console.WriteLine($"F1 Score: {metrics.F1Score:P2}");
Console.WriteLine($"AUC: {metrics.AreaUnderRocCurve:P2}");

The evaluation metrics help determine whether the model generalizes well to text it hasn't seen during training.

7. Predict sentiment for new text

Create a prediction engine:

PredictionEngine<SentimentData, SentimentPrediction> engine =
    mlContext.Model.CreatePredictionEngine<
        SentimentData, SentimentPrediction>(model);

var input = new SentimentData
{
    Text = "The product is fantastic and works perfectly!"
};

SentimentPrediction prediction = engine.Predict(input);

Console.WriteLine(
    $"Sentiment: {(prediction.Prediction ? "Positive" : "Negative")}");

Console.WriteLine($"Probability: {prediction.Probability:P2}");

A complete minimal flow is therefore:

CSV training data
       ↓
Load with ML.NET
       ↓
Train/test split
       ↓
FeaturizeText
       ↓
Binary classifier
       ↓
Train
       ↓
Evaluate
       ↓
Predict new text

Improving the model

For a real application, the quality and diversity of your labeled training data are often more important than simply changing algorithms. You can improve the system by:

Using a substantially larger labeled dataset.
Including examples representative of your actual users and domain.
Handling class imbalance.
Comparing several ML.NET classifiers.
Performing cross-validation and hyperparameter tuning.
Cleaning or normalizing text when appropriate.
Tracking precision, recall, F1, and AUC rather than relying only on accuracy.

For production workloads, you can also save the trained model:

mlContext.Model.Save(
    model,
    split.TrainSet.Schema,
    "sentiment-model.zip");

and load it later:

DataViewSchema schema;

ITransformer model = mlContext.Model.Load(
    "sentiment-model.zip",
    out schema);

This lets you train the model separately from the application that performs predictions.

MindStick Training

Markdown for AI

A clean, structured version of this page for AI assistants and LLMs.