What Is SentiSignal and How It Works

Market Sentiment, Decoded by AI

SentiSignal is an AI-powered market sentiment intelligence platform that continuously monitors financial news and scores every article for sentiment — so you can see at a glance whether the media is bullish, bearish, or neutral on the assets you care about.

Instead of manually reading hundreds of headlines, SentiSignal ingests articles from 180+ news sources, runs them through a dual sentiment engine, and presents the results as interactive charts, feeds, and analytical tools.

What Does SentiSignal Monitor?

The platform covers three major asset classes with a total of 89+ instruments:

• Crypto — 20 symbols including BTC, ETH, SOL, ADA, XRP, DOGE, AVAX, LINK, DOT, MATIC, and more
• Commodities — 30 symbols including GOLD, SILVER, OIL_BRENT, OIL_WTI, PLATINUM, PALLADIUM, COPPER, NATURAL_GAS, WHEAT, CORN, and more
• Forex — 39 currency pairs including EUR/USD, GBP/USD, USD/JPY, AUD/USD, USD/CHF, and more

The Dual Sentiment Engine

Every article is scored by two independent engines:

1. LLM (Large Language Model) — Google Gemma-3, an AI language model that reads the full article and returns structured sentiment scores with reasoning. This is the primary, high-accuracy engine.
2. VADER (Valence Aware Dictionary and sEntiment Reasoner) — A rule-based NLP algorithm that scores text instantly using a sentiment lexicon. Free, fast, and useful as a secondary signal.

Both engines produce a score on the same scale: -1.0 (extremely bearish) to +1.0 (extremely bullish), with 0.0 representing neutral sentiment.

How the Pipeline Works

From the moment a news article is published to the moment you see its sentiment on your dashboard, it passes through a multi-stage pipeline:

1. News Ingestion — RSS feeds and source APIs are polled continuously, pulling in new articles from 180+ sources.
2. Algorithmic Prefilter — A fast keyword-based check determines whether the article is likely relevant to any tracked asset. This rejects roughly 40-50% of irrelevant content before expensive AI processing.
3. LLM Relevance Check — The title and snippet are sent to the LLM for a quick relevance assessment.
4. Full Content Crawl — Relevant articles have their full text fetched and cleaned.
5. Deep LLM Analysis — The full content (up to 6,000 characters) is analyzed. The LLM returns per-symbol sentiment scores, quality, credibility, and importance scores, plus a written reasoning.
6. VADER Analysis — In parallel, VADER produces its own compound sentiment score.

Performance and Caching

SentiSignal uses a 3-layer cache to deliver fast page loads:

• Browser cache — Static assets and short-lived data cached locally
• Cloudflare CDN — Edge-cached responses served from the nearest data center
• Redis — Server-side cache for computed aggregates (sentiment averages, chart data)
• PostgreSQL — The durable source of truth for all articles, scores, and metadata

Key Features at a Glance

• Sentiment Charts — Interactive overlay charts showing sentiment vs. price over time. Learn more in How to Read the Sentiment vs Price Chart.
• SentiLab Experiments — Statistical tools like Correlation Sweep and Sentiment Divergence to test whether sentiment predicts price.
• AI Chart Analysis — Automated written analysis combining chart patterns with sentiment data.
• Market Briefs — AI-generated daily summaries for each asset class.
• Source Leaderboards — See which news sources are the most accurate, credible, and predictive.

Why It Matters

Traditional charting tools show you what the price did. SentiSignal adds the why — the narrative context that drove sentiment shifts. By combining both views, you can identify whether current sentiment is leading or lagging price, whether the news flow is abnormally bullish or bearish, and whether specific sources consistently provide early signals.

Ready to get started? Continue to How to Set Up Your Dashboard and Follow Assets.