Skip to content

samples

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

samples

README.md

SentimentAnalyzer Samples

This folder contains sample applications demonstrating the capabilities of the SentimentAnalyzer packages.

📦 Packages Overview

Package Size Engine Languages Accuracy Best For
SentimentAnalyzer.Core <1 MB VADER (Lexicon) English ~75% WASM, Edge, IoT
SentimentAnalyzer.Onnx ~18 MB TinyBERT (Transformer) English 94.2% Mobile, Desktop
SentimentAnalyzer.Onnx.Multilingual ~5 MB + 270 MB* DistilBERT 104 languages ~88% Server, Multilingual

* Model downloaded on first use, cached for offline use

🎮 Sample Applications

1. Console Demo (SentimentAnalyzer.Samples.Console)

Interactive console application with a beautiful terminal UI (powered by Spectre.Console).

Features:

  • Quick demo comparing all 3 engines side-by-side
  • Interactive mode to analyze your own text
  • Built-in performance comparison tool
  • Package information display

Run it:

cd samples/SentimentAnalyzer.Samples.Console
dotnet run

2. Benchmarks (SentimentAnalyzer.Samples.Benchmarks)

Professional-grade benchmarks using BenchmarkDotNet to measure:

  • Latency: Time per prediction (short, medium, long texts)
  • Throughput: Predictions per second
  • Memory: Allocations per operation
  • Cold Start: Initialization time

Run it:

cd samples/SentimentAnalyzer.Samples.Benchmarks
dotnet run -c Release

Sample Results:

Engine Short Text Medium Text Throughput
VADER ~0.01ms ~0.05ms ~100,000/sec
TinyBERT ~5ms ~8ms ~150/sec
DistilBERT ~15ms ~25ms ~50/sec

(Actual results will vary based on hardware)

3. Blazor WebAssembly (SentimentAnalyzer.Samples.Blazor)

A fully offline sentiment analyzer running entirely in the browser via WebAssembly.

Features:

  • Real-time sentiment analysis
  • No server required - 100% client-side
  • Privacy-first: data never leaves the browser
  • Uses VADER engine (perfect for WASM due to small size)

Run it:

cd samples/SentimentAnalyzer.Samples.Blazor
dotnet run

Then open https://localhost:5001

Note: Currently uses VADER engine only. ONNX engines require native dependencies not yet compatible with Blazor WASM.

4. MAUI Cross-Platform (SentimentAnalyzer.Samples.Maui)

A native cross-platform sentiment analyzer for Windows, macOS, iOS, and Android.

Features:

  • Beautiful native UI
  • On-device AI inference
  • Uses TinyBERT for better accuracy (falls back to VADER if needed)
  • Works offline after installation
  • CommunityToolkit.Mvvm for clean architecture

Run it:

cd samples/SentimentAnalyzer.Samples.Maui

# Windows
dotnet run -f net10.0-windows10.0.19041.0

# macOS
dotnet run -f net10.0-maccatalyst

# Android (requires Android SDK)
dotnet run -f net10.0-android

# iOS (requires macOS + Xcode)
dotnet run -f net10.0-ios

🚀 Quick Start

  1. Clone the repository:

    git clone https://github.com/arafattehsin/SentimentAnalyzer.git
cd SentimentAnalyzer

  2. Restore and build:

    dotnet restore
dotnet build

  3. Run your preferred sample:

    # Console (easiest to start with)
dotnet run --project samples/SentimentAnalyzer.Samples.Console

# Benchmarks
dotnet run -c Release --project samples/SentimentAnalyzer.Samples.Benchmarks

# Blazor
dotnet run --project samples/SentimentAnalyzer.Samples.Blazor

📊 Why Offline Sentiment Analysis?

Feature Cloud APIs SentimentAnalyzer
Latency 100-500ms <1ms - 25ms
Privacy Data sent to server 100% on-device
Offline ❌ Requires internet ✅ Works anywhere
Cost Per-request pricing Free after install
Edge/IoT ❌ Limited ✅ Perfect fit
WASM ❌ Not possible ✅ Browser-native

🔧 Requirements

  • .NET 10.0 SDK or later
  • For MAUI: appropriate platform SDK (Windows SDK, Xcode, Android SDK)
  • For Benchmarks: Run in Release mode for accurate results

📝 License

MIT License - See LICENSE for details.