Unit RR, Part G: Optimizations

Accumulator:

• Sort query terms on frequency.
• Process low frequency terms first.
• Stop / prune:
  • no later term can (is likely to) change top n documents

Other:

• Screening of feedback queries, by:
  • user
  • statistics from relevant retrieved
• Space reduction -> faster lookup:
  • hashing if order is not important
• Tune to collection characteristics via parameters.