Information Retrieval Performance Measurement Using Extrapolated Precision William C. Dimm DESI VI June 8, 2015
Comparing Precision-Recall Curves
What if we only know one point?
Comparing Precision-Recall Points
F 1 Comparison: Wrong Conclusion!
F 1 Depends Strongly on Recall
F 1 Contours
Danger Zone
Want Contours Like P-R Curves
Less Cutting Across Contours
How to Quantify?
Precision's Relationship with Cost ● Precision is meaningful – inversely proportional to number of docs to review: n = ρNR/P
Extrapolated Precision
X is Fairly Independent of R
Precision-Recall Curve Math
Actual Precision-Recall Curves
Actual Probability Curves
Model Probability vs Actual
Model Precision vs Actual
Extrapolation Limitations ● P < 0.99 ● R < 0.99 ● P >= 2ρ / (1 + ρ + R*(1 - ρ))
Summary ● Proportionality dictates recall – Need performance measure less sensitive to recall ● Extrapolate precision-recall point to target recall level using model curves ● Model precision-recall curves are constant performance contours ● When close to target recall, performance measure inversely proportional to review cost