Authority Control in the 21st Century: An Invitational Conference

Authority Control Profile of a Large Academic Research Library Database

Table of Contents

• Background
• Results of the Survey
  • Profile of the Sample
  • Name Headings in the Sample
  • Frequency of Sample Headings in the Database
  • Conclusions
• Subject Headings
  • Topics of Interest
  • Complexity of Subject Headings
  • Frequency of Occurrence
  • Verification Against LCSH
  • Conclusions

Background

• Planning for automated authority control component for LIAS, Penn State's local system
• Presentation and article by Jennifer Younger
  • Concept of utility in authority control
  • Maximizing the impact of authority work
  • Correlation between disciplines and the way information is likely to be retrieved
  • Do humanists and scientists search online catalogs differently?
• Need for research

Research Plan: Two Studies

• Survey of the content of the online catalog [preliminary results reported here]
• Examination of user behavior related to use of the online catalog [to be conducted later in 1996]

Basic Question: Is there a relationship between subject matter and the distribution of types of headings or the frequency of occurrence of individual headings?

The Database Survey

The Sample

• A random sample of 449 records
• Sampled from records added to the online catalog 1983 through 1995
• All formats included
• In-process records excluded
• Expected to be representative of current cataloging

The "Questionnaire"

• A stripped-down MARC record
  • Included title and all access points
• Data collected on print-outs of records
• 68 questions asked for each record
• Data gathered on call number, language, names, subjects, and series

Sample Validation

• Call number breakdown of sample compared with October 1995 call number data for the entire Penn State database
• Almost an exact match

Disciplines

Based on Library of Congress call numbers

Humanities	A-B (except BF), M-N, P
History/Geography	C-F, G
Social Sciences	BF, GN-GV, H-L, U, V
Science/Technology	GA-GF, Q-T

Note: Z's assigned to discipline categories based on subject content of item.

Results of the Survey

Profile of the Sample

Number of Records by Language

English	382	85%
German	22	5%
Spanish	12	3%
French	11	2%
Other	22	5%
TOTAL	449	100%

• English accounted for an overwhelming majority of records in the sample
• Chart 1: 15K

Number of Records by Discipline

Humanities	141	31.4%
History/Geography	48	10.7%
Social Sciences	142	31.6%
Science & Technology	118	26.3%
TOTAL	449	100%

• Humanities, Social Sciences, and Science are more or less balanced.
• Chart 2: 16K

Number of Headings Per Record: Names

None	7	1.5%
1	289	64.4%
2	105	23.4%
3-9	47	10.5%
10+	1	0.2%
TOTAL	449	100%

• The overwhelming majority (almost 88%) of the records had 1 or 2 names.
• Only 7 records had no name headings.
• Chart 3: 14K

Number of Headings Per Record: Subjects

None	45	10.0%
1	145	32.3%
2	117	26.1%
3	78	17.4%
4	40	8.9%
5+	24	5.3%
TOTAL	449	100%

• 45 records (10%) had no subject headings.
• The overwhelming majority (almost 78%) of the records had 1 to 3 subject headings.
• Chart 4: 15K

Number of Headings Per Record: Series

• 63% of the records had no series.

Name Headings in the Sample

There were a total of 667 name headings in the sample.

Number of Names by Type of Name

Personal	505	75.7%
Corporate	144	21.6%
Conference	18	2.7%
TOTAL	667	100%

• Chart 5: 15K

Number of Names by Discipline: Personal Names

Humanities	188	37.2%
History	45	8.9%
Social Sciences	140	27.7%
Science	132	26.2%
TOTAL	505	100%

• 37% of the personal name headings were in Humanities (only 31% of the records in the sample were in Humanities)
• Chart 6: 15K

Number of Names by Discipline: Corporate Names

Humanities	19	13.2%
History	16	11.1%
Social Sciences	61	42.4%
Science	48	33.3%
TOTAL	144	100%

• 42% of the corporate name headings were in Social Sciences (only 32% of the records in the sample were in Social Sciences).
• 33% of the corporate name headings were in Science (only 26% of the records in the sample were in Science).
• In contrast, only 13% of the corporate name headings were in Humanities (31% of the records in the sample were in Humanities).
• Chart 7: 15K

Number of Names by Discipline: Conference Names

Humanities	0	0.0%
History	1	5.5%
Social Sciences	4	22.2%
Science	13	72.3%
TOTAL	18	100%

• Most of the conference names were in Science.
• However, out of the 667 name headings in the sample, there were only 18 conference names.
• Chart 8: 15K

Frequency of Sample Headings in the Database

Each name heading in the sample was searched in the Penn State online catalog (about 1.75 million records), and the number of records with that heading were counted.

Number of Names by Type of Name

Frequency	1	2-5	6-10	11-99	100+
Personal Names	148	184	66	96	11
Corporate/Conference Names	21	22	10	36	55

• Most personal names occur 5 times or less in the database.
• Most corporate and conference names occur 10 times or more.
• Chart 9: 17K

Number of Personal Names by Discipline

Frequency	1	2-5	6-10	11-99	100+
Humanities	52	55	25	45	11
History	10	18	9	8	0
Social Sciences	37	60	17	26	0
Science	49	51	15	17	0

• Personal names are most likely to occur frequently in Humanities.
• Chart 10: 18K

Number of Corporate and Conference Names by Discipline

Frequency	1	2-5	6-10	11-99	100+
Humanities	4	5	3	4	3
History	4	1	1	1	10
Social Sciences	12	2	6	17	28
Science	10	20	1	16	14

• Corporate and conference names are most likely to occur frequently in Social Sciences.
• Chart 11: 18K

Conclusions

• Type of Name
  • Personal names are concentrated in Humanities
  • Corporate and Conference names are concentrated in Social Sciences and Science.
• Frequency of Occurrence
  • Personal names are less likely to occur frequently.
  • Corporate and conference names are more likely to occur frequently.

Subject Headings

Topics of Interest

• Structure of subject headings:
  • How complex are the headings?
  • How many subdivisions do they contain?
• Verification of subject headings:
  • How many headings are in LCSH?
  • What is not in LCSH?

Complexity of Subject Headings

Of 906 subject headings in the sample:

• 31% had only one subdivision ($a)
• 42% had two subdivisions
• 27% had three or more subdivisions

Frequency of Occurrence

Looking at each subdivision in the sample independent of the rest of the heading and searching each subdivision in the entire Penn State online catalog:

• 49% of the main headings ($a) occurred in more than 100 headings in the catalog
• 92% of the topical subdivisions ($x) occurred in more than 100 headings
• 85% of the chronological subdivisions ($y) occurred in more than 100 headings
• 93% of the geographic subdivisions ($z) occurred in more than 100 headings

Verification Against LCSH

Complete Headings

• Total of 816 LC subject headings (650 or 651)
• 365 of these (45%) could be verified completely in LCSH
• 451 (55%) could not be verified completely
• Almost always, this was because one or more subdivisions were not in LCSH

Topical Subdivisions ($x)

• The 816 subject headings contained 473 topical subdivisions
• 149 of these subdivisions were in LCSH, 324 were not
• Of these 324 unverified subdivisions, there were 122 unique subdivisions
• Of these 122 unverified topical subdivisions, 111 are free-floating subdivisions
• 75 of these appear in H1095
• Conclusion: Subdivision records for the headings in H1095 would greatly enhance the odds of verification

Chronological Subdivisions ($y)

• The 816 subject headings contained 51 chronological subdivisions
• 39 of these subdivisions were in LCSH, 12 were not
• Of these 12 unverified subdivisions, there were only 2 unique subdivisions: 19th century and 20th century

Geographic Subdivisions ($z)

• The 816 subject headings contained 264 geographic subdivisions
• Only 33 of these subdivisions were in LCSH, 231 were not
• Of these 231 unverified subdivisions, there were 74 unique subdivisions
• Of the 74 unique geographic subdivisions:
  • 3 were names of continents
  • 35 were names of countries
  • 11 were names of states
• 49 of the subdivisions belong to these three categories
• Conclusion: If subdivision records were created for these three categories, the odds of verification would be greatly increased
• Note: For all three categories, the form of the subdivision is the same as the form of the main heading, so the subdivision records could be created automatically.

Impact of Subdivision Records on Our Sample

• Using just LCSH:
  • 365 headings (45%) were verified
  • 451 headings (55%) were not
• Using LCSH plus the subdivision records recommended
  • 705 headings (86.5%) were verified
  • Only 111 headings (13.5%) were not

Conclusions

• Subject headings are complex and hierarchical.
• LCSH does not contain a majority of the complete subject headings appearing in bibliographic records.
• Subdivision records can dramatically increase odds of verification.

Copyright © 1996 Christine Avery, Mary Ann Itoga, John C. Attig

Return to Proceedings Home Page

Advanced Search | Careers at OCLC | Feedback | Privacy Policy | ISO 9001 Certificate | ©2003 OCLC Online Computer Library Center, Inc.