Recall measures how many relevant or positive items a system successfully finds. A system with 90% recall retrieves 9 of 10 relevant items and misses one. section 3.13
Retrieving legal cases
A law firm searching a large case database may prioritize recall so relevant cases are not missed. But retrieving almost every case can push recall near 100% while producing many false positives and very low precision. section 3.13
Recall is not useful by itself. Retrieving very little can produce high precision but poor recall, while retrieving nearly everything does the reverse. F-measure combines precision and recall, though equal weighting may not match business priorities. section 3.13 section 3.15
F = 2(P R)/(P + R)
Use the F-measure equation to see how precision and recall combine into a single score. section 3.13
