What Is SentiLab and What It's For

SentiLab: Your Quantitative Research Platform

SentiLab is SentiSignal's built-in data science experimentation platform. While the main asset pages give you charts and analytical tools, SentiLab lets you go deeper — running statistical analyses that quantify the relationship between news sentiment and price movements.

It transforms SentiSignal from a sentiment dashboard into a quantitative research tool.

The screenshot shows the Quick Analysis panel. At the top, three cards let you choose the Analysis Type: Correlation Sweep (highlighted with blue border), Source Predictability, and Narrative Clustering — each with a short description. Below that: Asset Category buttons (Crypto, Metals, Energy, Agriculture, Forex), an Asset Symbol dropdown showing "Bitcoin (BTC) (85 421 news)", a Time Range (days) field set to 90, and a blue "Analyze" button. A hint below reads: "Quick mode uses 90 days + predictive news filter for best results."

Three Analysis Types

SentiLab offers three distinct analysis methods, each answering a different question:

1. Correlation Sweep

Finds the optimal time lag where sentiment best predicts price. Uses Cross-Correlation Function (CCF) with False Discovery Rate (FDR) correction for statistical rigor. Answers: "Does sentiment lead price for this asset, and by how many hours?"

Detailed lesson: Understanding Correlation Sweep

2. Source Predictability

Ranks news sources by their predictive power. For each source, it tests whether their articles tend to lead or lag price movements. Answers: "Which news sources should I follow most closely for early signals?"

Detailed lesson: Source Predictability Analysis

3. Narrative Clustering

Detects media convergence events — moments when many sources suddenly report on the same topic simultaneously — and analyzes their impact on price. Answers: "Does herd media behavior create contrarian trading opportunities?"

Detailed lesson: Narrative Clustering Analysis

Getting Started: Quick Analysis

Access SentiLab at /lab (requires a free account). The Quick Analysis mode gets you started in seconds:

1. Select your analysis type (Correlation Sweep, Source Predictability, or Narrative Clustering).
2. Choose your asset (e.g., Bitcoin, Gold, EUR/USD).
3. Set a time range (e.g., last 90 days).
4. Click Analyze.

Results appear within a few seconds, displaying charts, tables, and interpretation text.

Advanced Configuration

For more refined analyses, expand the Advanced Configuration panel. This gives you access to:

• Quality filters — minimum article quality score (0–1).
• Credibility filters — minimum source credibility score (0–1).
• Lag range — how many hours forward/backward to test (Standard ±24h, Extended ±72h, Weekly ±168h).
• Keywords — filter to articles containing specific terms.
• Intent type — filter by article intent (future_oriented, past_oriented, etc.).

These filters are covered in depth in the Advanced Filters lesson.

Saving and Publishing Experiments

Every analysis can be saved as an experiment — storing both the configuration and results for later reference. You can return to a saved experiment to review its findings or re-run it with updated data.

You can also publish an experiment to share it publicly on the /experiments page. Published experiments go through a quality gate: the system evaluates the experiment using an LLM and assigns a score out of 100. Only experiments scoring ≥ 60/100 can be published. This ensures the public experiment feed maintains a baseline quality standard.

Why This Matters

Most crypto sentiment tools stop at "here's the sentiment, here's the price." SentiLab goes further by letting you quantify whether sentiment actually matters for a specific asset, which sources are predictive, and whether media herd behavior creates exploitable patterns. It's the difference between looking at a chart and doing data science on it.