Recommendation Algorithms

Implements scoring in Kotlin -> compiled to WASM/JS for Cloud usage.

Activity Scoring

Components:

• Keyword Jaccard similarity
• Type weighting

Formula:

score = 0.6 * keyword_jaccard + 0.4 * type_similarity

Event Scoring

Factors:

• Base keyword similarity
• Event kind weight (map)
• Novelty (inverse of occurrence count)
• Recency (timestamp decay)

Diversity

Applied as penalty if top-N contain similar titles (>80% token overlap).

Performance

• Pre-computed keyword sets
• Single pass weighting
• WASM linear memory arrays for scores

Data Normalization

Before scoring, articles/events normalized (lowercase, trimmed, stripped HTML).