SentimentOracle

SentimentOracle applies transformer-based NLP models to millions of social media posts, forum threads, and news articles every hour to produce a real-time sentiment score for 500+ crypto tokens. Each score ranges from -1 (extreme fear) to +1 (extreme greed) and is accompanied by volume and velocity metrics. Beyond the headline score, the API exposes topic clusters, influential post highlights, and anomaly alerts when sentiment for a token diverges sharply from its 7-day moving average. These signals are valuable for timing entries, gauging narrative momentum, and detecting coordinated hype campaigns. Data sources include X/Twitter, Reddit, Telegram public channels, Discord announcement feeds, and a curated set of crypto news outlets. Models are fine-tuned weekly on labelled crypto-specific corpora to maintain accuracy.