Automated batch authority control processing has advanced significantly since the 1980s when it was first introduced. In the early days there were few vendors and their procedures evolved from controlling headings in the production of printed book and microform catalogs.
Because the demands of computerized catalogs far exceed those of card and book catalogs, LTI's authority control procedures have been designed to meet the more stringent authority requirements of the online environment. Given the enormous variation that exists among vendors in the thoroughness and quality of their authority control processing, LTI takes the position that customers have the right to know in advance what is being purchased. LTI takes the guesswork out of authority control by guaranteeing a minimum link rate of library headings to authorized headings for US libraries following national cataloging standards.
What's Included In LTI's Authority Control Processing
LTI's authority control processing includes filing indicator fixes; pre-processing fixes and MARC updates; machine linkage of headings to authority records, including limited manual review of headings that remained unlinked following machine processing; optional full manual review of unlinked headings; final re-linking; and writing of bibliographic and linked authority records to separate files for transmittal to the library or its local system vendor.
Each step in the authority control process is designed to maximize authority record links and to eliminate incorrect links. Where the full heading in the catalog record cannot be validated or linked to an authority record, LTI attempts to link portions of the heading. Two types of batch authority control are available: limited manual review and full manual review.
What is Limited Review?
Limited review compares, validates, and updates library headings by computer programs and fix tables. Formerly referred to at LTI as "machine" or "machine-only" processing, all limited review batch authority jobs receive some level of editor review. High frequency unlinked headings are always examined. Single-occurrence unlinked headings are reviewed if they fall into one of many categories of "problem" headings. Problem headings are software-identified and include a range of data and content designator errors. Examples are detection of a music subfield code ($m,$r,$o) without the presence of a title subfield, a series that lacks a title, presence of digits in subfield $a of a personal name, a birth date that falls after a death date, etc.
In limited review processing, LTI guarantees that 95% or more of the customer's controlled headings will be linked to either an LC or LTI authorized heading. This assumes, of course, that the library has tried to adhere to national cataloging standards. Typically, about 93% of the headings link to LC authority records and 4% to LTI authority records. If less than 95% of the library's controlled headings link to an authority record, LTI performs further review to raise the overall heading link rate to 95% at no cost to the customer.
In addition to a guaranteed link rate, distinguishing features of LTI's authority control service include:
While there is no requirement to use either continuing service (AEX or AUP), for clients that do use them, there will never be a need to do a batch re-authorization of the entire database. Moreover, controlled headings always remain in synch with national cataloging practices. In effect, every AUP run represents a total reauthorization of all the library's bibliographic records
In most library databases, we find that with minimal processing about 70% of the normalized headings will match exactly with a 1XX field, or a 4XX pointing to a 1XX, in an LC name or subject authority record. In other words, provided the authority control vendor does not introduce problems that destroy links, with minimal effort any vendor should be able to link seven out of ten library headings to LC authorized headings.
What distinguishes superior authority control from mediocre processing is what the authority control vendor does with the 30% of the headings that have failed to link initially to an authorized heading. An important factor in LTI's high link ratio of library headings to LC authority record headings is the large number of headings linked via machine based on years of collecting and analyzing several hundred million bibliographic headings.
How LTI Gets Such a High Link Ratio
LTI uses a combination of fuzzy logic and "fix and loop" routines to manipulate and repeatedly test headings against authorized headings. Depending on the type of heading (personal name, uniform title, topical subject, etc.), unlinked headings go through a cycle of routines designed to make high precision full or partial links to authorized headings. Often headings can be validated fully with an authorized heading only after one or more fixes are made to subordinate units.
LTI's fix procedures and tables are based on well defined rules of what constitutes a valid and invalid heading, as well as on empirical analysis of errors that have appeared in library headings processed by LTI over the past 25 years. For example, the subject subdivisions "United States" and "Description and travel" are each represented by hundreds of variations in subject subdivision fix tables. Headings that cannot be validated against an LC authority record are checked against 4.4 million LTI-created cross-references from incorrect headings to the authorized LC heading. For instance, our files include over one hundred cross-references (i.e., in this case variant forms and misspellings) for the composer "Tchaikovsky." If an LC link cannot be made, the heading is matched against LTI authority records. Matching against LTI's enhanced LC cross-reference file occurs after LC authority record links have been made but prior to LTI authority record linking.
Diacritics and special characters are retained in the match key. LTI's normalized headings also retain delimiters and subfield codes. This permits the construction of programs that check for links with and without all subfields in the heading. For example, if a personal name heading that contains $c and $d fails to link with an LC authority record, the library's heading is tested for linkage without the subfield $c, and the $d is also checked for permissible variations. For example, the birth dates $db.1952 and $d1952- are considered equivalent.
Wherever possible we have replaced editor review of unlinked headings with computer-based correction and matching techniques. Writing programs to optimize linkages while avoiding mismatches is a balancing act that requires expertise both in cataloging and in manipulating headings in MARC records. We also recognize that every backfile authorization project benefits from some level of editor review. LTI's experienced librarians have devoted years to analyzing why headings do not link, what processing is needed to achieve links, and what changes are needed to bring unlinked headings into conformity with current cataloging standards. LC weekly authority record files are scrutinized for changing patterns of usage. In this way LTI is able to implement current LC cataloging decisions sometimes well in advance of notification in LC's Weekly Lists or Cataloging Service Bulletin.
What About Full Manual Review?
Given LTI’s high level of linking with limited review, full review is not considered a cost effective approach for most libraries. If during machine or limited review processing 96% of the library's subject headings are validated against authorized headings, the potential of increasing the link rate by several percentage points probably will not merit the additional expense. Cost benefit is another way to evaluate machine versus manual review processing. [These figures are based on statistics from LTI processing and are not applicable to other authority control vendors.] On average, for bibliographic records taken from OCLC or other distributors of LC MARC records, roughly one percent of the library's controlled headings will be changed during full manual review that would not have been changed during limited review processing. For example, assuming a database of 200,000 catalog records with each record containing 3.8 controlled headings-i.e., a total of 760,000 controlled headings-full manual review processing might change about 7,600 headings that would not be changed during limited review processing. To achieve this 1% link rate improvement, the cost of a library's authority control project will be more than double that of a limited review job. For most libraries, selecting limited review processing is an opportunity to save money without paying a huge penalty in quality.