claimbuster
**ClaimBuster** is an automated system developed at the University of Texas at Arlington that identifies **check-worthy factual claims** in text — the first and crucial step in the automated fact-checking pipeline. It scores sentences based on their likelihood of containing important, verifiable factual claims.
**How ClaimBuster Works**
- **Input**: Takes text input — a debate transcript, speech, news article, or any text containing potential claims.
- **Scoring**: Each sentence receives a **check-worthiness score** from 0 to 1, indicating how likely it is to contain a factual claim that is worth verifying.
- **Ranking**: Sentences are ranked by their scores, allowing fact-checkers to focus on the most important claims first.
- **Classification**: Sentences are classified into categories — **Non-Factual Sentence (NFS)**, **Unimportant Factual Sentence (UFS)**, and **Check-Worthy Factual Sentence (CFS)**.
**Technology**
- **Training Data**: Trained on thousands of sentences from US presidential debates, political speeches, and other public discourse, labeled by professional fact-checkers.
- **Features**: Uses linguistic features (named entities, numbers, sentiment), structural features (sentence position, length), and contextual features (topic, speaker).
- **Models**: Evolved from SVM classifiers to transformer-based models (BERT fine-tuning) for better performance.
**Applications**
- **Live Debate Monitoring**: Process debate transcripts in real-time to highlight check-worthy claims as they are made.
- **News Analysis**: Scan news articles to identify factual claims that should be verified.
- **Social Media Monitoring**: Flag viral posts containing check-worthy claims for fact-checker review.
- **Fact-Checker Workflow**: Prioritize which claims to check first based on check-worthiness scores.
**API and Access**
- **ClaimBuster API**: Publicly available API that scores text for check-worthiness.
- **Integration**: Can be integrated into newsroom workflows, social media monitoring tools, and fact-checking platforms.
**Significance**
ClaimBuster addresses a fundamental bottleneck in fact-checking — **there are far more claims made than fact-checkers can verify**. By automatically identifying the most important claims, it helps fact-checkers allocate their limited time to the claims that matter most.
ClaimBuster represents an important step toward **scalable fact-checking** — it doesn't verify claims itself but ensures that human fact-checkers focus on what matters.