TruthLens — AI Misinformation & Claim Verifier

Two Realms. One Truth.

Next-generation AI fact-checking engine. Cross-referencing public web intelligence, NLI stance classification, multi-source consensus scoring, and token explainability.

Live Web Retrieval & Fact Archives Multi-Source Consensus Gauge NLI Stance Verification
Benchmark Tests:
Interactive Verification Pipeline

How Buzz AI Verifies Claims

Explore the multi-tier inspection pipeline engineered for institutional fact-checking accuracy.

Stage 01 — Scrape & Ingest

Multi-Format Content Scraping

Accepts raw text or website URLs. Trafilatura extracts main article body text while stripping boilerplate HTML, navigation, and advertisement noise.

Asynchronous scraping via Celery task queue
Automatic boilerplate and nav removal
SHA-256 prediction hashing for Redis caching
Architectural AI Foundations

Engineered for Precision & Speed

Every layer of Buzz is built for institutional verification performance and forensic transparency.

DistilBERT Transformer Model

Fine-tuned transformer neural network exported to ONNX Runtime format. Evaluates contextual semantic patterns across headlines and full text bodies in under 50ms.

Latency: < 45msQuantization: INT8 ONNX

spaCy Stylometry

Analyzes part-of-speech distributions, clickbait syntax, sentence sentiment polarization, and lexical density.

Stylometric NLP Features

TF-IDF N-Gram Baseline

Statistical term frequency analysis cross-references unigrams and bigrams against verified datasets.

Logistic Regression Baseline

SHA-256 Redis Caching & SlowAPI Limits

Every submission is hashed with SHA-256 for instant Redis cache lookups, preventing redundant GPU inference. SlowAPI protects endpoints from automated abuse.

Cache Hits: Instant ResponseRate Limit: Enforced
Interactive Claim Showcase

Examine Forensic Evidence

Select an investigative sample below to inspect real-time token attribution and credibility scoring.

Selected Article AnalysisReal (94% Confidence)

International research teams publish peer-reviewed findings confirming breakthroughs in room-temperature energy storage materials.

SHAP Token Attribution Matrix
International (+0.2)research (+0.3)teams (+0.1)publish (+0.4)peer-reviewed (+0.5)findings (+0.2)