Releases: arafattehsin/SentimentAnalyzer
v3.0.0 - Multi-Engine Architecture
🎉 Major Release: Multi-Engine Sentiment Analysis
This release introduces a complete architectural overhaul with 4 distinct NuGet packages:
New Packages
🔹 SentimentAnalyzer.Core - VADER-based lightweight engine (<1 MB, ~75% accuracy)
🔹 SentimentAnalyzer.Onnx - TinyBERT transformer (~18 MB, 94.2% accuracy)
🔹 SentimentAnalyzer.Onnx.Multilingual - DistilBERT multilingual (104 languages, ~88% accuracy)
Legacy Package
🔹 SentimentAnalyzer - ML.NET-based (backward compatible with v2.x)
Key Features
- Fluent Builder API: Choose your engine with \SentimentAnalyzer.CreateBuilder().UseVader().Build()\
- Modular Design: Use only what you need
- Cross-Platform: .NET Standard 2.0, .NET 8, .NET 10
- Sample Apps: Console, Blazor WASM, MAUI, Benchmarks
Model Files
The DistilBERT multilingual ONNX model is attached to this release. It's automatically downloaded on first use by SentimentAnalyzer.Onnx.Multilingual.
Note: This is a correctly exported ONNX model. The official HuggingFace ONNX export has inverted predictions.
Documentation
See the README for full documentation, comparison tables, and migration guide.
All Changes
See CHANGELOG.md for detailed release notes.