Addendum to Guest Blog: Talking Turkey

Gordon V. Cormack and Maura R. Grossman

+ Continuous Active Learning [acttoplite]
× Simple Active Learning [untoplite]
✳ Simple Passive Learning (keyword seed) [pastoplite]
☐ Simple Passive Learning (random training) [pasranlite]

Results as one pdf

Results as separate pdfs

Note: For ease of comparison with our SIGIR study, the graphs above portray the use of training-set sizes (for simple active and simple passive learning) that minimize 75% recall effort. For easy comparison with Webber's results, graphs portraying training-set sizes that minimize 80% recall effort (denoted cfr80 by Webber) are available here.