APPENDIX C
TRANSACTION LOG RECORD FORMAT
This appendix describes the types of data captured for each search query in the transaction log record.
TRANSACTION LOG RECORD FORMAT
| Field Name | Description |
| User's password | Each user's password is captured during logon process |
| Logon date and time | Logon date and time (to the nearest second) is supplied by the system |
| Full search query | The full search statement entered by the user |
| Parsed search query | The full search statement is preparsed. Three pieces of data are recorded: 1) stemmed query terms; 2) term id's; and 3) collection weight of each search term |
| Normalized call numbers | Normalized call numbers in user-selected clusters are recorded in the log file |
Top Ranked Clusters
(Up to 20 cluster records can be displayed)
Query id number Query id number assigned by the system
Cluster id number Each cluster record that best represents all the records in a given cluster is assigned a unique number in advance
Cluster rank The rank of each retrieved cluster record (1st, 2d, etc.) among all the retrieved cluster records
Cluster weight Cluster weight determines how closely each cluster record matches the query entered by the user
Action taken by the user Code indicating whether the user displayed the cluster record and whether he or she selected it as relevant or not
Normalized call number Classification number that best represents all the bibliographic records in a given cluster
Topical information Broad topical information taken from the Library of Congress Classification (LCC) scheme (e.g., Bibliography Libraries . . .)
Number of records Number of bibliographic records represented by a given cluster
Subject headings The most frequently encountered three Library of Congress subject headings attached to the bibliographic records under a given cluster (along with their frequencies)
(continued)
TRANSACTION LOG RECORD FORMAT (Continued)
| Field Name | Description |
Top Ranked Bibliographic Records
(Up to 20 bibliographic records can be displayed)
Query id number Query id number assigned by the system
Record id number ID number of record displayed
Record rank The rank of each retrieved bibliographic record (1st, 2d, etc.) among all the retrieved bibliographic records
Record weight The weight of bibliographic record that determines how closely the record matches the query entered by the user
Feedback iteration number Whether the record is retrieved during original search or relevance feedback search (users can perform relevance feedback searches more than once for the same query)
Action taken by the user Code indicating whether the user displayed the bibliographic record and whether he or she selected it as relevant or not
Bibliographic data The full MARC data (author, title, imprint, subject headings, etc.) for each retrieved bibliographic record including local call numbers
Completion time Date and time supplied by the system once the user completes the search
Records Retrieved During Relevance Feedback Searches
(Up to 20 records can be displayed)
Should the user opt for a relevance feedback search after the first retrieval results, CHESHIRE revises the original query based on the user's relevance judgments and retrieves more bibliographic records. The same type of information as given above (under Top Ranked Bibliographic Records) is displayed for each bibliographic record retrieved during relevance feedback searches. The user can continue relevance feedback searches as many times as desired. Starting and ending times are supplied for each relevance feedback cycle.