How It Works

How market sentiment analysis works

Market prices don't move in a vacuum. News, narratives, and the emotional reaction of market participants precede and follow every significant price move. Sentiment analysis quantifies that reaction — so you can see it, not just feel it.

What is market sentiment?

Market sentiment is the collective attitude of market participants toward an asset at any given moment. When more news is positive and analysts are bullish, sentiment is high. When fear dominates the narrative, sentiment drops — often before the price does.

Sentiment doesn't replace price analysis — it adds a layer. A rising price with falling sentiment can signal a trend about to reverse. A price crash with stabilizing sentiment can indicate capitulation is near. SentiSignal makes this relationship visible and quantified.

How we measure it: LLM + VADER

🤖 LLM analysis (primary)

A large language model reads each article in full context — understanding nuance, regulatory implications, irony, and mixed signals that word-counting methods miss. It scores sentiment from -1.0 (very negative) to +1.0 (very positive) and assigns a relevance score to the asset.

  • • Understands context, tone, and implicit meaning
  • • Catches hedging, sarcasm, implicit bearishness
  • • Scores relevance alongside sentiment
  • • Handles complex regulatory and macro language

VADER cross-check (supplementary)

VADER (Valence Aware Dictionary and sEntiment Reasoner) is a rule-based NLP model used to cross-validate LLM scores and catch any systematic drift. Its deterministic output acts as a consistency anchor — especially useful for high-volume news events.

  • • Deterministic — same input always yields same score
  • • Zero hallucination risk — lexicon-grounded
  • • Processes thousands of articles per minute
  • • Flags anomalies in LLM output for review
Why LLM first? Financial news is dense with nuance — a headline that reads bullish on the surface can be bearish in context. LLMs understand that. VADER ensures the output stays consistent and auditable at scale.

Reading the sentiment score

Raw sentiment scores (-1.0 to +1.0) are converted to a 1–5 star display for quick reading. The score shown is a 7-day rolling average — smoothing out single-article spikes and showing the underlying trend.

★★★★★Very Bullish+0.6 → +1.0
★★★★☆Bullish+0.2 → +0.6
★★★☆☆Neutral-0.2 → +0.2
★★☆☆☆Bearish-0.6 → -0.2
★☆☆☆☆Very Bearish-1.0 → -0.6

The 7-day average prevents a single viral negative headline from flipping the score overnight. Use shorter timeframes in the chart to see real-time swings.

XRPRipple · live 30-day chart
Open full page →

The sentiment line moves independently of price — watch for divergences where sentiment leads price by hours or days.

What's on an asset page

Every asset page (e.g. Bitcoin, Gold, EUR/USD) combines all the platform's data in one view.

Module 1

Sentiment + Price Chart

The main chart overlays the asset price (candlestick or line) with a sentiment trend line. When sentiment leads price, you see it here — the blue line moving before the candles respond.

Module 2

News Event Markers

Significant news articles are pinned directly on the chart at the time of publication. Click any marker to see the headline, sentiment score, and how the price reacted in the following hours.

Module 3

News Feed

A filtered, AI-ranked stream of news relevant to this specific asset. Each article shows its sentiment contribution, source credibility badge, and a 1-sentence AI summary.

Module 4

Market Briefs

Clustered summaries of related news topics. When multiple sources converge on the same story, a brief consolidates them into a 200-word signal instead of 20 duplicate articles.

Module 5

AI Articles

Deeper analysis pieces generated by the AI pipeline when a topic reaches critical mass. Grounded in real data: price context, source breakdown, sentiment trend.

Module 6

AI Chart Analysis

One-click AI interpretation of the current chart view. Select a time window and get a plain-language explanation of what the price and sentiment data shows.

Module 7

Source Leaderboard

Every news source is scored by predictive track record. Sources that consistently publish articles before price moves rank higher. You can filter your feed by source quality.

Source credibility & the Leaderboard

Not all news sources are equal. A tweet from a micro-influencer and a Reuters report both contain text — but they carry different weight. SentiSignal scores each source by:

Predictive accuracy
Does this source publish before price moves or after?
Consistency
Does the sentiment it expresses track with eventual outcomes?
Reach & uniqueness
Is it a primary source or a re-aggregator of other content?

The Source Leaderboard for each category (Crypto, Metals, Forex) ranks sources by these metrics. You can filter your news feed to only show high-credibility sources.

See it for yourself

Create a free account to access the full sentiment dashboard, personalized news feed, and AI analysis.