Methodology & Enrichment Reference

Every enrichment field is produced by deterministic, auditable algorithms — VADER lexicon scoring, keyword/regex pattern matching, heuristic formulas, and conditional LLM augmentation. No black boxes. Every score is reproducible from source code.

Methodology Types

VADER LexiconKeyword/RegexHeuristic FormulaLLM (conditional)Regex PatternDirect PassthroughHeuristicEmotion LexiconSliding-Window VADERSHA-256 Hash

Why This Matters

Deterministic by DefaultOver 95% of enrichment fields use lexicon matching, regex, or heuristic formulas — fully reproducible from source code.

LLM is OptionalGROQ Llama 3.3 70B runs only when GROQ_API_KEY is set. Without it, all core sentiment, emotion, quality, and classification fields still populate from deterministic algorithms.

Quality-GatedEvery enrichment passes 30+ validation checks. Corrections are logged in quality_corrections. Contradictions (e.g. positive sentiment + anger emotion) are automatically resolved.

SocialINTEL

Fetching datasets…

Why This Matters

Deterministic by DefaultOver 95% of enrichment fields use lexicon matching, regex, or heuristic formulas — fully reproducible from source code.

LLM is OptionalGROQ Llama 3.3 70B runs only when GROQ_API_KEY is set. Without it, all core sentiment, emotion, quality, and classification fields still populate from deterministic algorithms.

Quality-GatedEvery enrichment passes 30+ validation checks. Corrections are logged in quality_corrections. Contradictions (e.g. positive sentiment + anger emotion) are automatically resolved.

Methodology & Enrichment Reference

Methodology Types

Sentiment Analysis

8-Emotion Model

Text Quality & Content Scoring

Content Classification & Topic Detection

Engagement & Reach

Signal Detection (FUD, Bot, Misinformation)

Financial & Crypto Signals

Domain-Specific Extraction

Jobs & Career Fields

Language Detection

Aspect-Based Sentiment

Computed Composite Scores

LLM-Augmented Fields

Quality Control & Corrections

Why This Matters

Methodology & Enrichment Reference

Methodology Types

Sentiment Analysis

8-Emotion Model

Text Quality & Content Scoring

Content Classification & Topic Detection

Engagement & Reach

Signal Detection (FUD, Bot, Misinformation)

Financial & Crypto Signals

Domain-Specific Extraction

Jobs & Career Fields

Language Detection

Aspect-Based Sentiment

Computed Composite Scores

LLM-Augmented Fields

Quality Control & Corrections

Why This Matters