Format of Search Results to NACSIS
(enhanced on Feb. 6, 1999)
(modified on Nov.29, 1998)
0001 0 gakkai-0000000001 1 9999 ntc1
0001 0 gakkai-0000000002 2 9998 ntc1
0001 0 gakkai-0000000006 3 9997 ntc1
0001 0 gakkai-0000000004 4 9997 ntc1
0001 0 gakkai-0000000005 5 9995 ntc1
0002 0 gakkai-0000000001 1 9999 ntc1
0002 0 gakkai-0000000002 2 9998 ntc1
0002 0 gakkai-0000000006 2 9998 ntc1
The search result file which will be sent to NACSIS should follow the format
Topic-ID Dummy-field Document-ID Rank Similarity-value Run-ID
- Elements in a line are segmented by a TAB.
- The list must be sorted numerically by Topic-ID in ascending order.
- The Dummy-field is not used for evaluation, usually use '0' for the field.
- Document-ID is ACCN in each document record.
- Rank is represented as an integer. But the evaluation program ignores the
- Similarity-value, or brief-value is a score that each system produced to
present the similarity of the document to the topic. A float. The Similarity-value
is assumed to be higher for the documents to be retrieved in the higher ranks.
programme will sort and rank the retrieved document by
Similarity-value, then Document-ID.
Similarity-value is mandatory for evaluation. If the system does not produce
such score, please assign some value in the field.
- Run-ID is a tag given to the run by the submitter(participant group).
Run-ID should begin with the Group's ID, which was specified in
the application form by the participant group.
- When a participant group
submits more than one result
file, add sequential number to the Run-ID to differenciate each result file.
For example, when the 'ntc' group submits three result files, the Run-ID's
are 'ntc1', 'ntc2', and 'ntc3'. This number can be different
from the priority of the judgment, which will be specified later
in the web form of the "Search result submission form"
- The Run-ID must be the same on every lines in a result file.
- The result file on the NTCIR's ftp site in a file named by the Run-ID.