i.e. 10% of relevant documents retrieved, 20% recall, 30% recall, etc. to give a set of 10 recall 

precision figures), Figure 10. 

Recall Precision 

10 67.3 

20 65.9 

30 59.2 

40 45.3 

50 36.7 

60 33.3 

70 21.9 

80 19.7 

90 15.3 

100 12.1 

average precision 

37.67 

Figure 10: Example recall and precision figures  

With a test collection, the recall precision (RP) figures for each query are averaged to form a single set 

of recall precision figures

13

. The averaged RP figures are often averaged across the recall points to give 

a single value   the average precision value, Figure 10. 

RP figures are often represented graphically. Figure 11 shows an example of a recall precision graph 

drawn from the RP figures of two systems on the same test collection. As the line for System 1 is 

entirely above the line for System 2 we can infer that System 1 is better than System 2. 

Precision

100

90

80

70

60

System 1

50

System 2

40

30

20

10

0

0

10

20

30

40

50

60

70

80

90 100

Recall

Figure 11: Example RP graphs 

13

Interpolation measures are necessary for queries whose recall levels differ from the standard, e.g. the example in 

Figure 10 is based on 10 recall levels, any query with a number of relevant documents different from a multiple of 

ten. Interpolation is often used to calculate a 0% recall figure to give an 11pt recall precision table. 

 16