Table 4

11-point precision diagram. This example shows a query that is submitted to two different IR systems (IR1 and IR2), which are based on the same collection of 20 documents. Both IR1 and IR2 rank all 20 documents, of which 10 are relevant. However, IR1 ranks the relevant documents higher on average than does IR2. The mean average precision for IR1 = 0.79 and for IR2 = 0.40. The recall and precision curves for IR1 and IR2 are shown in figure 2.

Ranking by IR1
Ranking by IR2

Ranking
Doc
Relevant
Recall
Precision
Ranking
Doc
Relevant
Recall
Precision

1
d1
yes
0.10
1.00
1
d1
no
0.00
0.00
2
d2
yes
0.20
1.00
2
d2
no
0.00
0.00
3
d3
yes
0.30
1.00
3
d3
no
0.00
0.00
4
d4
no
0.30
0.75
4
d4
no
0.00
0.00
5
d5
yes
0.40
0.80
5
d5
no
0.00
0.00
6
d6
no
0.40
0.67
6
d6
yes
0.10
0.17
7
d7
yes
0.50
0.71
7
d7
yes
0.20
0.29
8
d8
no
0.50
0.63
8
d8
yes
0.30
0.38
9
d9
yes
0.60
0.67
9
d9
no
0.30
0.33
10
d10
no
.060
0.60
10
d10
yes
0.40
0.40
11
d11
yes
0.70
0.64
11
d11
no
0.40
0.36
12
d12
yes
0.80
0.67
12
d12
yes
0.50
0.42
13
d13
yes
0.90
0.69
13
d13
no
0.50
0.38
14
d14
no
0.90
0.64
14
d14
yes
0.60
0.43
15
d15
yes
1.00
0.67
15
d15
no
0.60
0.40
16
d16
no
1.00
0.63
16
d16
yes
0.70
0.44
17
d17
no
1.00
0.59
17
d17
yes
0.80
0.47
18
d18
no
1.00
0.56
18
d18
yes
0.90
0.50
19
d19
no
1.00
0.52
19
d19
no
0.90
0.47
20
d20
no
1.00
0.50
20
d20
yes
1.00
0.50

Zhou et al. Journal of Biomedical Discovery and Collaboration 2006 1:2   doi:10.1186/1747-5333-1-2