How does it Work?

• Scatter/Gather clusters similar documents

• Clustering posits similarity measure

  • we used vector space model
  • relevance/similarity via dot product
  • works surprisingly well

• Scatter/Gather has 2 phases

  • preprocessing: 1Gb/day (can be incremental)
  • interaction: 5 seconds, regardless of corpus size

Previous slide

Next slide

Back to first slide

View graphic version