1	MARC Content Designation and Utilization An Empirical Investigation of Metadata Utilization by Library Catalogers The MARC Content Designation Utilization Project William E. Moen <wemoen@unt.edu> School of Library and Information Sciences Texas Center for Digital Knowledge University of North Texas
2	Metadata – Rules & Practice Library catalogers create metadata – bibliographic records Follow cataloging rules and other standards to create the bibliographic data Encode the bibliographic data into MARC records MARC – communications format and metadata scheme Approximately 2,000 structures for encoding data
3	Richness of MARC
4	What do catalogers use? Given the cataloging rules… Given the detailed structuring of bibliographic data in MARC records… Given training of the catalogers… Given local policies and practices… What can we learn by examining a large set of MARC bibliographic records?
5	The MCDU Project MARC Content Designation Utilization Provide empirical evidence of catalogers’ use of MARC content designation Identify commonly used elements of bibliographic records Contribute to community discussion about core elements in MARC bibliographic records Explore the evolution of MARC content designation Develop research approach to understand the factors influencing levels of MARC content designation use
6	Metadata Record as Artifact Metadata creation as process Decisions by metadata creators Influenced by… ? Artifact: something created by humans for some practical or utilitarian purpose Artifact reflects decisions, policies… Artifact reflects metadata utilization decisions Decisions to use or not use available metadata elements
7	Metadata Utilization Analysis Developing a re-usable methodology Developing software tools and database design for storing large datasets Identifying questions to be answered Determining methods for analyzing data to address questions Compiling results
8	Dataset and preparation 56,177,383 MARC 21 Bibliographic Records from OCLC WorldCat Decomposed the records to store in MySQL Parsing Tool 82 hours to process and load records 295 GB final database size (with indexing) Structuring of decomposed records align with analytical questions
9	Analytical questions What is the average length of the records? What is the average length of the records? What is the frequency of types of records? What is the status of the record? What is the frequency of encoding levels? What is the frequency of the descriptive cataloging forms? What are the total occurrences of all control and data fields? What are the total occurrences of each control and data field? What are the total occurrences of all subfields? What are the total occurrences of each subfield? What percentage of records contains at least one occurrence of each control and data field? What percentage of records contains at least one occurrence of each subfield?
10	Additional data preparation Analysis required determining frequency counts by format of material (ten) Concern about significant differences in patterns of utilization between Library of Congress and OCLC member cataloging Partitioned decomposed data into 20 databases Based on source of cataloging Based on format of material
11	Querying database Q08: What is the distribution of records by Type of Record? Natural Language query: Interrogate table LEADER, field 06_RecType. Count occurrence of each possible value of field 06_RecType and group by 06_RecType. SQL query: select `06_RecType`, count(`06_RecType`), count(`06_RecType`)/(select count(`06_RecType`) from `Leader`) from `Leader` group by `06_RecType`
12	Questions to Answer Two categories of questions General profile of the dataset What is the distribution of records by Type of Record? What is the distribution of records by Length of Record? Occurrences of content designation structures, What is the number of total occurrences of all control and data fields and how many unique field tags are used? In how many and in what percentage of records is each unique field/subfield combination used at least once?
13	Occurrence Summary
14	Example Results 7,595,887 LC-created records in dataset Type of Record: Book, Pamphlets, and Printed Sheets Total number of unique fields: 167 Number of fields accounting for 80% of occurrences: 14 fields (8.3%) Number of fields accounting for 90% of occurrences: 21 fields (12.6%) Approximately 110 fields occur in less than 1% of all records. [Note: Fields are cataloger-supplied, not system-supplied]
15
16	Making sense of numbers Frequency counts provide raw but informative data Threshold – concept to delineate a change in trend in utilization Determining commonly occurring elements Comparing to recommended core records Comparing to recommendations for national level records Comparing the FRBR user tasks data
17	Implications Empirical basis for decisions about core elements in a metadata scheme Profiling repositories of metadata for aggregators Organizations can use methodologies and tools to analyze local utilization levels Contributions to changes in cataloging rules, practices, policies, and standards
18	References MARC Content Designation Utilization Project http://www.mcdu.unt.edu/ Assessing Metadata Utilization: An Analysis of MARC Content Designation Use http://www.unt.edu/wmoen/publications/MARCPaper_Final2003pdf.pdf Format Content Designation Analysis: Set Profiling and Analysis Queries http://www.mcdu.unt.edu/wp-content/FANSetProfilingAnalysisQuerieswemFinal20Dec05.pdf