Taxonomic Shift Detection

Overview

Detect when a content producer migrates from one interest subgroup to another over time by comparing windowed category distributions, applying change point detection to the distribution time series, and identifying which historical period best represents the current voice. The core insight: interest shifts are measurable changes in category distribution shape, not just changes in the most-frequent category -- a user who gradually spreads across three new domains has shifted just as meaningfully as one who abruptly swaps a primary interest.

REQUIRED BACKGROUND: You MUST have a category-mapped corpus before using this skill. Use taxonomic-interest-classification to map content items to categories first. This skill operates on the temporal evolution of those category distributions, not on raw text.

When to Use

• Corpus of categorized content spans a meaningful time period (months to years) and you need to detect interest migrations
• Need to identify the "current voice" period -- which era of the timeline best represents the user's present interests
• Investigating whether observed category changes are genuine shifts or temporary exploration
• Feeding era-segmented profiles into downstream archetype assignment or voice modeling
• Detecting life transitions, professional pivots, or topic drift from behavioral data patterns