How to measure Diversity of Search Results
summary
Provides a concrete, repeatable measurement: compute Shannon entropy over result-source categories to score diversity, implemented as a notebook (Apache Zeppelin) and a shared Google Sheet. Demonstrates real numbers: Library of Congress Mark Twain top-24 results score 0.78 vs 1.81 when comparing “all items” to “available online.” Shows integration with relevance checks using Cupid and a practical category-prediction threshold (examples 0.7 and 2.5) so auto-faceting helps when entropy indicates a clear dominant category.
AI-written summary · about
chapters (6)
- 0:00Introduction to diversity
- 5:07Precision vs recall in search
- 10:37Measuring search result diversity
- 18:48Shannon entropy in practice
- 24:12Category prediction techniques
- 28:10Q&A on diversity metrics
related talks
19mPopping the Filter Bubble
45mWords in Space: A Visual Exploration of Distance, Documents, and Distributions...
35mDiversity and Data: Cases in the Music Industry
01h 04mDemonstration workshop freedom index app
59mEverybody lies
47mAssessing the quality of a clustering
30mWords in Space
50mSerialrank: Spectral Ranking using Seriation