distilbert-political-tweets
| Property | Value |
|---|---|
| License | LGPL-3.0 |
| Framework | PyTorch, TensorFlow |
| Dataset | senator-tweets |
| Accuracy | 90.76% |
What is distilbert-political-tweets?
distilbert-political-tweets is a specialized natural language processing model fine-tuned on US Senator tweets from the first year of the Biden Administration. Built upon the DistilBERT architecture, this model excels at classifying text as having either Democratic or Republican sentiment, achieving an impressive 90.76% accuracy and 91.17% F1 score.
Implementation Details
The model is based on distilbert-base-uncased and was trained on 99,693 tweets with a near-balanced distribution (51.6% Democrat, 48.4% Republican). Training utilized the Adam optimizer with a learning rate of 5e-5 over 5 epochs, implemented using Transformers 4.16.2 and TensorFlow 2.8.0.
- Balanced dataset of nearly 100,000 tweets
- Float32 training precision
- Optimized for short-text political sentiment analysis
Core Capabilities
- Binary classification of political sentiment (Democratic/Republican)
- Optimized for tweet-length content
- High accuracy on contemporary political discourse
- Real-time inference capability
Frequently Asked Questions
Q: What makes this model unique?
This model specializes in modern American political discourse analysis, trained on actual senator tweets from 2021, making it particularly relevant for contemporary political sentiment analysis.
Q: What are the recommended use cases?
The model is best suited for analyzing short political statements, social media content, and tweet-length political discourse. It's important to note that accuracy may decrease with longer text passages.
