Stage 02 — Processing

Process Data Into Signals

Raw media goes in. Structured signals come out. This is what happens between extraction and insight.

[Network Processing]

//001

SIMILARITY

Overlap calculated

  • Follower overlap analysis
  • Following pattern matching
  • Mutual connection scoring
  • Pairwise relationship weights
//002

GRAPH BUILD

Structure assembled

  • Directed edge construction
  • Connection matrix encoding
  • Optimized graph representation
  • Edge weight normalization
//003

ALGORITHMS

Intelligence extracted

  • Community detection
  • Influence scoring
  • Force-directed layout
  • 2D/3D position assignment

Network Signals Output

  • Community membership assignments
  • Influence scores per account
  • Graph coordinates for visualization
  • Ready for Network Intelligence layer

[Media Processing]

//001

VISUAL

Frames decoded

  • Keyframe extraction
  • Scene composition
  • Object detection
  • Motion analysis
//002

AUDIO

Sound understood

  • Speech transcription
  • Audio classification
  • Semantic embedding
  • Language detection
//003

TEXT

Meaning mapped

  • Caption analysis
  • Hashtag extraction
  • Entity recognition
  • Semantic encoding
//004

PERFORMANCE

Numbers measured

  • Performance metrics
  • Engagement benchmarks
  • View tracking
  • Interaction rates

Unified Signal Output

  • Visual, audio, and text signals merged into unified representation
  • Every piece of content becomes searchable and comparable
  • Signals feed forward into Network, Content, and Conversation analysis
  • No manual tagging — intelligence extracted automatically

[Comment Processing]

//001

CLASSIFY

Text understood

  • Neural text encoding
  • High-dimensional representation
  • Batch processing pipeline
  • Semantic space mapping
//002

SENTIMENT

Scores calculated

  • Calibrated sentiment scoring
  • Benchmark comparison
  • Semantic similarity analysis
  • 5-point Likert scale output
//003

DEDUPLICATION

Noise removed

  • Text normalization
  • Semantic clustering
  • Canonical comment selection
  • Spam and bot filtering

Comment Signals Output

  • Sentiment scores per comment
  • Deduplicated canonical set
  • Topic and entity extractions
  • Ready for Conversation Intelligence layer

What These Signals Enable

Processed signals feed into three intelligence layers