CRANV2
Aslib Cranfield Research Project: Factors Determining the Performance of Indexing Systems: Volume 2
Conclusions
chapter
Cyril Cleverdon
Michael Keen
Cranfield
An investigation supported by a grant to Aslib by the National Science Foundation.
Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government.
- 261 -
Document Index
Number Terms
1 ADx
2 AIBIDlX
3 ACDlx
4 BlX
5 BCDAlx
6 ClX
7 BClX
8 DlX
9 ABCDx
10 AlX
11 BDAlx
12 ClX
13 AB1ClX
14 D x
15 B[OCRerr] Dx
16 Dx
17 C[OCRerr]1 BI x
18 ClX
19 Ax
20 BlX
Assuming the search now is for any coordination of (A + A1), (B +
(C + C1) and (D + D1), the retrieval at different coordination levels willBlbe)'
as follows
Coordination Level No. of Documents Retrieved
2 (Document 5, 9)
8 (Document 5. 9. 2, 3, 11. 13. 15. 17)
10 (Document 5. 9. 2, 3, 11, 13. 15, 17.
1, 7)
20 (Document 5, 9, 2, 3. 11, 13. 15, 17,
1, 7, 4, 6, 8, 10, 12, 14, 16, 18, 20)
Again it is shown that the sum of the retrievals (40) equals the total
postings for the four groups of terms. Assume now that there were four
relevant documents, numbers 3, 7, 9 and 15. The performance in the two
cases would then be as follows
Coordin- Case A Case B
ation R N-R Recall Prec- R N-R
Level Recall Prec-
Ratio ision
Ratio ision
Ratio
4 1 0 25% 100% 1 1 25% 50%
3 2 1 .50% 66% 3 5 z5% 38%
2 3 3 75% 50% 4 6 100% 40%
1 4 6 100% 40% 4 16 100% 20%