Authorities Study Group Report
June 2, 1998

The Authorities Study Group was formed to make recommendations to the Library Technology Team for implementation of Unicorn's authority control features and to Library Administration for funding support.

The original study group, appointed in August 1997, was composed of:

Ramona Madewell, Chair
David Carpenter
Mary Charles Lasater
Sylvia Martin
Linda Tesar
Annette Williams
Pete Wilson

Suellen Stringer-Hye replaced Ramona Madewell as chair in Sept. 1997, when Ramona left Vanderbilt. Nancy Boggess-Korekach was added to the group at around the same time.

In order to meet its charge, the Study Group has met many times, sometimes around computers so that issues could be made clearer. Most of the group, along with other members of the staff, met with three SIRSI representatives in February to discuss various questions we had. This meeting was very useful.

II. The Charge

The Study Group was asked to:

• Review authority files and authorities processing needs relative to Acorn.
• Review Unicorn options for structuring authority files.
• Review options for automated updating of authority files.
• Understand all of the implications for how this work is to be done within the Acorn environment.
• Make recommendations to Library Technology relative to file structures and index and file processing schedules.
• Make recommendations to Library Administration about external services to be engaged for purposes of updating bibliographic records, authority records, and/or authority files.

Authority records on Acorn serve two purposes. The first and most important purpose is the generation of cross references that direct users to the forms of subjects, names and series titles that catalogers have used on bibliographic records. If a patron searches under a form which is not used in the bibliographic records, an authority record can provide a cross reference directing the patron to the authorized form. Authority records also generate cross references directing the user from one authorized heading to another, related authorized heading.

In some ways the cross reference structure that Unicorn builds from the authority file is more useful than what we had in our previous system. For example, Unicorn provides references to both broader and narrower subject headings, as shown here:

CATALOG LOOKUP BY CROSS REFERENCE

For topical and geo POPULATION BIOLOGY
1 search also under the broader term: BIOLOGY
2 search also under the narrower term: AGE-STRUCTURED POPULATIONS
3 search also under the narrower term: ANIMAL POPULATIONS
4 search also under the narrower term: BIOLOGICAL INVASIONS
5 search also under the narrower term: BIOTIC COMMUNITIES
6 search also under the narrower term: GROUP SELECTION EVOLUTION
7 search also under the narrower term: MICROBIAL POPULATIONS
8 search also under the narrower term: PLANT POPULATIONS
9 search also under: ECOLOGY

This is a detailed and fairly clear display that can help a patron find the subject heading he/she needs.

Unicorn does not handle name or uniform title references as well, however. These often turn out to be "blind" (they lead the patron to no records). The most common such case is this: a corporate or personal name reference shows up in a subject search result display, but the heading referred to has only been used as an author in the bibliographic records. The patron is directed to perform a subject search on the name, but when he/she executes the search he/she is told that there are "no entries found." (See Public Service Concerns about "Blind Reference" Displays in Acorn for further comment on this problem and its consequences.) SIRSI is working to correct this problem, but a fix will not come quickly. An enhancement request regarding this issue was among the top 20 requests from the 1998 Sirsi Development Survey. It asks that the "see also" cross references not display when no bibliographic records contain the heading referred to, and suggests that clarification of the "Sorry, no items were found" message would be an interim solution. Sirsi's response is that this request is "under investigation."

There are labor-intensive, non-standard ways in which we could "fix" the problems ourselves--by jerry-rigging authority records through recoding--but then we might have to undo all of that work when SIRSI comes up with their solution. In any case, we have barely enough staff to do standard authorities work, let alone this type of system-specific "correction," so the Study Group does not recommend that we undertake this type of project. The Study Group concurs with the aforementioned enhancement request in that changing the "Sorry" message to something clearer and more helpful to the patron may be the best solution at present.

The second purpose of authority records is to support cataloging. The authority record indicates which heading should be used in an access field for a person, corporate body, subject, uniform title or series. An authority record for a series title also indicates whether or not we trace the series, whether we classify items in the series separately or together, and whether or not we analyze the series (make separate records for each item in the series). Unicorn supports this function through the "browse authority" search and through automatic "unauthorized" stamps on headings in the bibliographic records, which indicate that either a heading does not match the appropriate authority record or (more commonly) an authority record does not yet exist for the heading. Unicorn provides many options for locally configuring these operations. (See Authorities Workflow Notes for Resource Services and the Law and Biomedical Libraries for fuller detail on how authority work fits into cataloging workflows locally; see Using Unicorn's Authority Control Functions for fuller detail and discussion on those functions and the specific ways we implement them.)

Along with the blind references problem, the current lack of an operative MeSH (Medical Subject Headings) authority file in Acorn is one of our most pressing functionality concerns. MeSH is a thesaurus developed in the 1960s by the National Library of Medicine, and is used by the Biomedical Library to select subject headings for their bibliographic records (MeSH headings also appear on some non-Biomedical catalog records). Catalogers at Vanderbilt have wanted to load a MeSH file onto our library system for many years, to make the selection and authentication of MeSH subject headings easier. While planning for the conversion to Unicorn, we anticipated loading the MeSH authority records and set up an empty authority file for them. Later we discovered that this file was not "turned on." The few MeSH authority records that had been manually input into NOTIS had been converted incorrectly into the LCSUBJECT authority file, and the nonfunctional MeSH file will not accept newly input MeSH authority records.

Without a MeSH authority file in place, medical subject headings appearing in bibliographic records are not put through the authorization process that LC subject headings and names go through. Staff must check paper manuals for MeSH terms, and there is considerable room for human error. Additionally, the patrons are denied the cross reference structure that MeSH authority records would provide. This is a significant problem when a subject is obscure or an old, familiar subject heading is replaced by a new one.

For more information, see Overview of System Configuration Policies in Unicorn

When we migrated from NOTIS to Unicorn, our existing authority records were funneled into three authority files: LCNAMES, LCSUBJECTS, and LOCAL. (A MeSH file was created for anticipated MeSH authority records, but never became functional and is empty.) The categorization was based on record type. LCNAMES includes personal, corporate, conference, and geographic names and series titles. LCSUBJECTS contains only topical Library of Congress subject headings. (The Library of Congress itself treats most geographic names as subject headings rather than name headings; this discrepancy between LC practice and SIRSI implementation is theoretically problematic but has not really adversely affected cataloging work.)

The LOCAL file was created to contain an existing resource file of LC subject headings that had not been used on Acorn bibliographic records. This file was never updated since its loading in 1996. Because catalogers and authorities staff found it was more time-consuming to update one of these records (when the heading was actually used on a bibliographic record) than it was to simply bring in a new record from our Z39.50 connection to the Library of Congress, the Study Group recommended earlier this year that the nearly 130,000 records in the LOCAL file be deleted. The file was, in fact, deleted in early April. This decision had no impact on the OPAC but has speeded authority work.

As explained above, the Study Group would like to have a MeSH authority file added to the system. This would give us three separate authority files. (This would be different from the "MeSH button," a function that SIRSI told us about when we were in the process of choosing a new system. We have learned that this is simply a way to look up MeSH terms online--a thesaurus which is separate from the bibliographic and authority records. Using the MeSH button would be roughly equivalent to looking up subject headings in the print version of MeSH. It does not "authorize" headings or provide cross references to the public. The Group does not recommend that we implement this Unicorn option.)

Creation of a MeSH authority file does bring with it a potential problem. Currently, while a user can do a subject search limited to MeSH subject headings, they cannot search LC subject headings without also retrieving MeSH, local and other types of subject headings interfiled with the LC. LCSH cross references, provided by the LCSH authority file, are also displayed to the public in the results of this subject search. If, after the creation of a MeSH authority file, the public's subject search option continues to include MeSH headings as well as LCSH headings, it would seem appropriate to provide MESH cross references as well. This could result in considerable confusion for the patron, however, because a term which is an authorized heading in one subject heading system may not be valid in the other, and reference structures may conflict in other ways. A classic example of the incompatibility of the two systems is the "cancer vs. neoplasms" loop. A user might find these two cross references:

Neoplasms see CANCER
Cancer see NEOPLASMS

Clearly this would be distressing to the patron.

This problem does not stand in the way of creating a MeSH file; there are ways to solve it. It would be possible to choose to post MeSH cross references only to the separate MeSH search, while letting the "combined" search stay as it is now, showing only LCSH cross references. However, the Study Group recommends that separate search options for LCSH subject headings and MeSH subject headings be provided, each with only its single appropriate cross reference structure included. (Other Vanderbilt catalogers, staff at other Sirsi libraries, and even Sirsi staff seem generally to concur with us in this.) We could still provide a combined subject search as well, or the remaining subject headings (the ones which are neither LCSH or MeSH) could be available only for keyword searching. (Prior to system migration, some catalogers left non-LCSH or MeSH subject headings on bibliographic records because of the enhanced keyword access that the terms provided; many are now eliminating these headings to avoid patron confusion.)

Before we migrated to Unicorn, the library used vendor support for updating the LC subject headings file. We wanted to use vendor support for keeping name (personal, corporate, conference, geographic) authority records up to date as well. However, we were unable to load these records from vendors, and had to rely on searching authority records on OCLC each time we used a heading to see if it had changed.

Since the migration, we have learned that the methods we previously used to update our subject file are not very workable in Unicorn, but we can batch-load all types of authority records via FTP. The Study Group has considered various authority control services offered by vendors and believes we could successfully use a vendor to keep our existing authority files, except for series authorities, up to date, and also to build a MeSH file. (We do not want to update series authority records automatically because local practice decisions are recorded in them, as well as many local changes to headings and references. Vendor-supplied updates to these records would override local changes. There are some local changes on other types of authority records, but we will attempt to retain them by not sending those records as part of the original file of authority records for the vendor to keep up to date.)

Basic proposal. In broad terms, there are two complementary ways in which we would like to engage a vendor to help us keep our authority files up to date. First, a vendor can be informed of what authority records are currently in our files; the vendor would then automatically send us, at regular intervals, the new versions of any of these authority records which are updated at the national level. These records would then be loaded into Acorn and would replace the outdated versions. (Marcive refers to this as a "notification service"; LTI calls their service "authority update processing.") The file kept by the vendor of authority records for which we wanted to receive updates would be dynamic. Authority records found by a vendor through processing our new bibliographic records (see below) would be added to it automatically. We would have to notify the vendor ourselves when we brought a new authority record into our database without actually cataloging anything new (as in retrospective authority work), or when we deleted an authority record because it no longer matched any headings in the bibliographic file (usually because the only item for which the authority record was required was withdrawn).

Second, all new bibliographic records added to the catalog can be sent via FTP to the vendor on a regular basis. All authority records needed to support the headings in these new bibliographic records would be sent back for us to load into Acorn.

Vendors can provide paper reports with their authority records, so that Vanderbilt staff can better assess the impact of loading the records into Acorn. These loads can eat up considerable amounts of system resources and time. Systems staff at Sirsi libraries are requesting that safeguards be built into Unicorn, so that large processing jobs could be halted when necessary or broken into more manageable pieces, but for now the problem remains. Also, changes and additions to the authority file require many hours of manual corrections of bibliographic records, especially in cases like subject heading splits, geographical name changes, or headings which have newly become geographically subdivisible. Paper reports help with this work.

The usefulness of paper reports was demonstrated recently in a special project involving Marcive. They performed authorities processing for us on a file of bibliographic records for the Chadwyck-Healey English Poetry Full-Text Database. In addition to producing a file of authority records for us to load, they corrected the bibliographic records, sending both files to us for loading. When Mary Charles examined the results of their work, she found authority records which apparently did not match any headings in bibliographic records, and other such inconsistencies. These were ultimately explained when we learned more about Marcive's policies (see below). However, in the process of resolving these problems, she found it necessary to print out a 200-page load report locally--a tedious process that we would not wish to repeat on a regular basis. If we do not have a vendor correct bibliographic records as part of their ongoing authority processing--and, judging from the poetry records instance, even if we do--we will routinely find authority records which do not exactly match any bibliographic headings in the database. In Sirsi there is no automated means of discovering such headings, so staff will have to work from paper lists; having the vendor supply these would save considerable time, effort and printer supplies.

Automated changes to bibliographic records? A vendor could also change headings in the bibliographic records sent to them, to match the authority records they provide. (They could also perform some other cleanup functions, such as correcting filing indicators for access fields beginning with articles.) Ideally, this would ensure a cleaner database and more accurate cataloging, but it would also cause some problems. While our records were "out" being massaged by the vendor, we would have to avoid making any changes to their local versions (since changes would be overlaid upon the return of the vendor-massaged records). In addition, because of the very complex array of origins for new cataloging--Resource Services, Law and Medical staff all contribute new records, and new records also come in through Government Documents tapeloads and other batch loads--organizing the process by which new bibliographic records are sent to the vendor and reloaded could become quite unwieldy. Finally, the reloading of the vendor-massaged bibliographic records would be a continuing burden on system resources and LTT staff.

As chance would have it, during the period of the Study Group's deliberations, the library engaged Marcive in a project that shed further light on the question of whether we should have a vendor correct our bibliographic records. This was the Chadwyck-Healey poetry records tapeload, mentioned above. Examining the results of Marcive's work, Mary Charles found that they, unexpectedly, had not corrected headings in bibliographic records which were "fuller" than the forms found in the matching authority records. This is their policy, and it probably works for some libraries. For us, however, it only means that a considerable number of bibliographic corrections would probably be left for authorities staff to perform even after the reloading of the massaged bibliographic records. This makes vendor correction of bibliographic records considerably less beneficial, without relieving us of any of its drawbacks.

Vendor possibilities. Members of the Study Group obtained information about four vendors supplying automated authority services: Library Technologies, Inc. (LTI); Marcive; OCLC; and WLN. Currently, OCLC, which acquired Blackwell's authorities processing services, is apparently reevaluating its operation and not taking customers. Little information was obtained about WLN's services, and members of the Study Group found them unresponsive when approached via email and at ALA.

More information was obtained about LTI and Marcive. Because Vanderbilt has used Marcive for a number of projects, one-time and ongoing, and found them to be efficient and professional, the Study Group admittedly found itself focusing on Marcive's offerings during much of its work. They offer services which match well the needs described above.

Some aspects of LTI's services concern us. For their processing, they create authority records of their own to supplement LC's authority file (they also add proprietary cross references to LC authority headings). They provide these authority records to their customers and use them in matching headings in bibliographic records. (These homemade headings allow them to guarantee that 95% of a customer's bibliographic headings will be linked to an authority record!) They have told us that they cannot do authority processing that disregards these non-LC name headings. To use their services would be effectively to outsource a considerable amount of intellectual work in authorities to an extent that we are not comfortable with, and to make our records potentially incompatible with national standards. Furthermore, as a NACO library, we should be creating new LC authorities ourselves when they are lacking.

LTI's promotional literature and a phone conversation with their president indicate that LTI can currently provide an authority updating service like the one described above (the second part of our "basic proposal") only to a library which has had LTI perform authorities processing on its entire database. (They call the service a "post-authority control" service; it depends on their retaining a file of the library's bibliographic records.) LTI provides new and changed authority records that match already existing headings in a library's bibliographic records, plus (depending on the level of service) either paper reports of bibliographic headings to be changed, or actual new bibliographic records to be loaded, replacing the older versions of the records. If, as we recommend below, we do not have the entire existing bibliographic database processed, this service will not be available. In any case, it would apparently utilize LTI's homemade authority records, which we wish to avoid.

We learned that LTI is currently planning to begin a simpler authority updating service, in which they would retain a file of authority records currently held by a library, and send them at intervals the new versions of any authorities which are changed at the national level. No corrections would be made to bibliographic records. This would be closer to what we are looking for. Apparently it will become available later this summer.

Cost comparisons. Detailed cost comparisons between vendors are awkward because of their different approaches to charging.

Estimates made by Marcive specifically for Vanderbilt include two different possible charges to start up their authority notification service. If we had them load a file of our complete authority records, they would charge around $3600; if we sent them only a file of the control numbers from the authority records via FTP, they would charge $1500. Thereafter, for a charge of $1350-1500 per year (depending on whether we wanted only electronic records, or both electronic records and paper reports), they would provide us with any changes to these authority records, on a monthly basis. We would be charged 5 cents per record to add or delete authority records from the file we want Marcive to track updates on, with a minimum batch charge of $10.00 (thus, we might want to save up authority record numbers until we'd met the minimum charge).

For authority processing of new bibliographic records "overnight over the Internet," Marcive would charge 9 cents per record. (It appears that if we did not use the notification service, this processing would cost 11 cents per record.) This charge includes changes to the bibliographic records, which we do not want, but we have not found that the charge can be reduced by refusing this part of the service. (Marcive's per-record charge for machine processing our whole database, or a section of it, in one-time batch mode would be considerably smaller.)

LTI's current authority update processing service does not appear applicable to us (see above). In a phone conversation with the president of LTI, we were given tentative cost figures for the simpler service they plan to begin later this year, which seems to be closer to what we want. It would involve an initial setup charge of $1000. If we wished to receive a file of changed authority records annually, we would pay $1000 per year. Semiannual processing runs would cost $800 per run, and quarterly runs would cost $600 per run. (These figures are considerably cheaper than those for their current authority update processing service.) No cost figures were obtained on what it would cost to add a record to, or delete one from, the file kept by LTI.

For ongoing authority processing on new cataloging--what they call their "authority express" service--LTI quotes a figure of 10 cents per bibliographic record, plus $10 per file of bibliographic records processed. This would cover LC name, LC subject and MeSH headings. Again, this price covers corrections to bibliographic records, which the Study Group does not want. LTI's promotional literature mentions a minimum charge of $500 for "any database processing work," but this does not apply to each "authority express" processing run.

We have not obtained much pricing information for WLN, beyond that they charge a profiling fee of $250 and charge 5 cents per record for "machine" processing. The per-record figure may or may not be accurate for ongoing authorities processing of new bibliographic records--it may apply only to one-time processing of a larger batch of records.

"Whole-database" option. It would be possible, though costly and somewhat cumbersome, to engage a vendor to perform authority work that would be much more extensive than that described above. We could send out our entire existing bibliographic database--not just new records as they are created--to a vendor. The vendor would identify and provide authority records--name, subject (LCSH and MeSH) and, if we wanted them, series--to match all headings in the bibliographic database. We would have to either eliminate our current authority files in favor of these newly provided authority records, or somehow reconcile the new and existing authority records. The vendor could also change incorrect or outdated headings in the bibliographic records to match the authority records they would provide, and perform miscellaneous bibliographic cleanup, such as correcting filing indicators.

This solution could, barring unforeseen complications, result in a considerably cleaner bibliographic database as well as up-to-date authority records. It would be somewhat expensive, however. We estimate that it would cost us around $37,125 to have Marcive perform these services on our approximately 1.5 million bibliographic records. In addition, we would have the heavy local burden of extracting our bibliographic records from the local system and then loading them back--probably in carefully managed segments--after the vendor had "massaged" them. We would have to delete our old authority files (thus losing all local authority records and local changes to authority records) and load the new vendor-supplied ones in--again, probably in segments--or figure out some other way to handle the two conflicting sets of authority files we would now have. Furthermore, as discussed above, we would have to refrain from changing the existing version of any bibliographic record while it was "out" with the vendor--this includes not just changes to the bibliographic record per se, but the addition of new copies, changes in acquisitions information, etc. This "freezing" of the records would be awkward and waste staff time.

Finally, we would be left, after the vendor processing, with a huge list of unmatched headings in bibliographic records. Every occurrence of a name that does not have a matching authority record would be included. Authors of Vanderbilt dissertations alone would take up a formidable pile of paper; huge numbers of names from older cataloging would also appear. (In addition, if Marcive did the processing, it appears that all name headings which were fuller forms of names that did have authority records would also be included, since Marcive does not correct these headings.) It is questionable whether we would ever have the time to work through these unmatched headings reports. When we had similar processing done before migrating to NOTIS, we received 14 boxes of such reports. No one ever had time to evaluate them, and they were eventually recycled.

It is true that, bibliographic corrections aside, whole-database processing would net us some new authority records needed to match currently unmatched headings. However, we did update our name authority file in the summer of 1996, and thus we suspect it is not badly out of date. We have been incorporating LCSH changes into our file manually and expect it is reasonably current.

Because of the costs and procedural implications, the members of the Study Group do not recommend the whole-database option.

What other libraries are doing. In an effort to compare our ideas on how to use vendor-supplied authority control services to the decisions of comparable libraries, members of the Study Group communicated with library staff at the University of Virginia and Emory University, two other Sirsi libraries, to learn how they are using Unicorn's authorities functions and what authority work they are having vendors do. As it turns out, neither library's approach appeals to us. The University of Virginia has put off the entire authorities issue--they are doing no online validation of their headings, and are waiting for the authorities workflow module, in hopes that it will make authorities work more practical for them. Clearly, they are not making use of any automated authorities services.

Emory has engaged LTI to create skeletal authority records for every heading on their bibliographic records not already covered by an LC or "LTI authorized" heading--including all used combinations of main heading and topical or even form subdivisions. For example, they would have an authority record for the LC subject heading "Psychology and literature," as well as LTI-made authority records for combinations like "Psychology and literature--Bibliography," "Psychology and literature--France," and so on. (These skeletal records are an option offered by LTI; most of their customers do not use this option.) This approach could be seen, superficially, as using Sirsi's heading validation capabilities to their fullest, but it would involve creating and storing immense, unwieldy numbers of authority records for very little practical purpose. (Several members of the Study Group think of this approach as "the tail wagging the dog.") Susan Bailey of Emory made no attempt to defend their approach to us, and indicated that she realized many libraries would not choose to do the same thing. She also indicated that the reports Emory receives from LTI are backlogged, waiting for her to have time to review the work. Neither Emory's approach nor Virginia's seems an example we would like to follow.

The MeSH special case. The Study Group devoted considerable time to the topic of MeSH. Currently, of course, we at least have authority files for names, LC subjects, and series. Because there currently is no MeSH authority file at all, MeSH is a special case. Unlike LC name and subject authority records, MeSH authority records are not available from either OCLC or the National Library of Medicine for downloading on a record-by-record basis. Therefore, batch loading is the only way we will be able to get MeSH authorities into the system. We have considered two possible sources for obtaining a beginning MeSH authority file--the National Library of Medicine, and Marcive. Details follow:

NLM solution: We would load the entire MeSH file, which includes records for main descriptors and for every main descriptor+allowable subheading category--about 500,000 records. The file is free from NLM under certain licensing restrictions; however, loading the entire file presents several problems:

• local cataloging records use only a fraction of the terms contained in the entire file
• unused headings would add more blind references to the catalog
• main heading+subheading authority records are not useful because the authorization process will be set up for main headings only
• loading such a large file could be problematic or necessitate loading in segments
• the cost of server space may not justify the return
• we would have to load annually a file containing all changes and additions to MeSH authority records

Marcive solutions: Elsewhere in this report the Study Group recommends sending all newly cataloged records to Marcive (including Biomedical records) for authority processing. If desired, Marcive will supply us with MeSH authority records as well as other kinds of authority records for this current cataloging. Thus getting MeSH authority records for new cataloging will not be a problem; it will happen "automatically."

However, this process alone will not supply us with MeSH authority records we need for the already existing backfile of Acorn records. The Study Group considered several options for obtaining MeSH authority records for all MeSH headings in the existing backfile, or in some subset thereof; they vary with respect to thoroughness and to cost. (One option that would be quite attractive--sending all Acorn bibliographic records with MeSH headings in them, whether they are for Medical materials or not, for processing--is apparently unavailable. As we understand it, there is no way to write a report that will identify and extract only that set of records.) Below, we estimate the cost of each option. (Marcive's estimates are for FTP delivery and include bibliographic record loading charge, authorities processing, output of bibliographic records to an FTP file, matching authority records, and their output to an FTP file. As stated elsewhere in this report, the Study Group does not recommend having a vendor make changes to our bibliographic records; the Biomedical representative to the Study Group does not see a great need for making an exception in the case of MeSH headings. Charges shown below are only for the processing associated with identifying headings which need authority records.)

Marcive option 1:Obtain MeSH authority records to match headings in all existing Acorn bibliographic records (i.e., in all libraries' records)

Pros:

• would build a complete authority file for all MeSH appearing in Acorn
• would authorize all MeSH headings in Acorn
• we could have the vendor perform other services on the bibliographic records for no extra cost (see note below)

Cons:

• 1.5 million records would have to be processed to identify the needed authority records
• records for some categories of material (e.g., periodicals, Biomedical's Historical Collections materials) probably don't really need to be processed because they will contain mostly rather broad subject headings, which will be duplicated on other records
• reloading the entire bibliographic file in Acorn (assuming we chose to do so) could be problematic or necessitate loading it in segments
• cost of server space may not justify the return

Note: As we understand it, if we did send all 1.5 million bibliographic records to the vendor just to get MeSH authority records for them, we would be charged the same fee as we would if we had the vendor do full authority processing and bibliographic cleanup on the records (see the discussion of "whole-database" vendor processing earlier in this section). Clearly we would be wasting money if we had the vendor do nothing but provide us with MeSH authority records; on the other hand, we have already shown some of the problems with more extensive processing.

Estimated cost: $37,125 (1.5 million records)

(The Study Group feels this cost is not justified.)

Marcive option 2: Obtain MeSH authority records matching headings in all existing Biomedical records.

Pros:

• would build a complete authority file for all MeSH appearing in all Biomedical records
• would authorize MeSH headings appearing in all Biomedical records
• would probably cover most MeSH headings occurring in non-Biomedical records as well

Cons:

• would require manual cleanup of non-Biomedical records containing unauthorized headings
• records for some categories of Biomedical material probably don't need to be processed (see option 1)
• would require asking Marcive for additional authority records for the backfile, as they are identified (piecemeal)

Estimated cost: $4287 (102,142 records)

(The Study Group feels that this would be an acceptable option, but slightly more expensive than necessary.)

Marcive option 3: Obtain MeSH authority records matching headings in all existing Biomedical records, except periodicals and Historical Collections.

Pros:

• would build an almost complete file for all MeSH appearing in Biomedical records
• would authorize most MeSH headings appearing in Biomedical records
• would probably cover most MeSH headings occurring in non-Biomedical and excluded Biomedical records as well
• would prevent some categories of material that contain broad MeSH headings from being processed unnecessarily

Cons:

• would require manual cleanup of non-Biomedical records and excluded Biomedical records containing unauthorized MeSH headings
• would require asking Marcive for additional authority records for the backfile, as they are identified (piecemeal)

Estimated cost: $3861 (91,743 records)

(The Study Group feels that the cost and benefits of this option are appropriate and recommends this option.)

Marcive option 4: Obtain MeSH authority records matching headings only in newly cataloged Biomedical records (i.e., do nothing with the backfile).

Pros:

• no cost beyond that of ongoing authority processing
• no loading of authority records other than those arriving through authority processing of new cataloging

Cons:

• loading only about 250 authority records per month will not build a useful MeSH authority file very quickly
• would leave the majority of MeSH headings unauthorized
• our local authority file would never be complete

(The Study Group feels that this option would be a poor solution.)

We will need to produce and send to a vendor a file of the existing authority record numbers that we want them to keep up to date. Additionally, we will need to start sending the vendor files of current cataloging so that they can supply authority records needed for these bibliographic records. Finally, we will need to keep track of authority records that we no longer need and report them to the vendor, so that the vendor can eliminate those records from the update service. If we do not report a "withdrawn" authority record, the vendor may send us an updated version of it that we do not need, potentially causing more blind references in the system. (System enhancements may eventually eliminate the need for this last step.) We will also need to keep track of authority records added to our files through retrospective work that does not involve current cataloging, and report these to the vendor; they will not be made part of our update service otherwise. Finally, of course, we will need to load the files received back from our vendor. The scheduling of these loads will be a concern.

Before any work can be done with the MeSH authority file, the policies governing this file must be established and the file must be "enabled." A separate MeSH thesaurus will need to be established to post cross references to the appropriate search index(es).

Currently there is no functioning test server in Acorn. A test server would allow us to experiment to discover the implications of certain actions we wish to take--for example, we would like to see how the posting of LCSH and MeSH cross references to the same index would affect users (see section IV). It would also allow us to experimentally discover how long the loading of authority records would take. Fortunately, action is being taken to provide a functioning test server in the very near future, so that Sirsi's Cataloging Workflows product can be tested. For authorities testing purposes as well as other purposes, the new test server should reflect current Acorn database structures and policies, and be updated as necessary to incorporate changes which are made to the main server.

Workflow changes. We envision two possible scenarios for changes to cataloging workflow resulting from vendor authority processing. In one scenario, we would include a bibliographic record in the file to send for processing as soon as the book was in hand and accepted for the collection--i.e., a firm order was received, or an approval was accepted. The matching authority records would be loaded, in most cases, prior to cataloging, eliminating "unauthorized" stamps from most new headings in the bibliographic records. If an "unauthorized" stamp did appear on a record for a speed or copy cataloging title, then the book would be sent to higher level staff for authority work and cataloging. This would allow authority checking to be done with the piece in hand, which is preferable to just working from a list of headings. This would also allow much of the authority work for the original cataloging stream to take place prior to the catalogers' receiving the titles, and would eliminate considerable authority file searching by original catalogers and support staff. Implementing this approach would require that we send new cataloging to Marcive frequently and load the authority records they send us quickly.

In a "next best" scenario, bibliographic records would be sent to a vendor at the point of cataloging, and authority records would be loaded prior to the running of "unauthorized headings" lists. This would streamline authority searching from lists (much of what did remain on lists would be special problems). This scenario would not help the original catalogers and their support staff avoid much authority checking, but would eliminate the need to load the records locally. RS, Law and Biomedical would all benefit from this approach-- authorities staff in all three locations now have to contend with unauthorized headings lists--but Divinity, which brings in all necessary new authority records at the point of cataloging, apparently would not.

A choice between these two scenarios will involve considerations of system capabilities and the workflow requirements of the various cataloging units; the Study Group feels there is not enough information at this point to choose between them. Either way, though, automated vendor support should save time for cataloging and authorities staff, resulting in more cataloging and project production. Catalogers who currently check headings in two authority files (Acorn's and either OCLC's or the Library of Congress's) should be able to rely on our local files. Finally, when Unicorn stamps a heading "unauthorized," we will know that it probably really is in error, rather than just not having a corresponding authority record, as is usually the case now.

Loading. We will receive two types of authority record files from a vendor regularly--files of updates to existing authority records, and files of new authority records needed for our current cataloging. Because of the size of our authority files, the Study Group recommends that we load the update files separately from the new cataloging files, at least until we are familiar with all the implications of this new service. Marcive provides update files monthly, whereas LTI apparently would provide them only quarterly in the service which they are currently planning; we would like to load updates at least monthly in order to keep our authority files and reference structures in agreement with headings on our bibliographic records. The files of new authority records derived from current cataloging should be obtained and loaded more frequently than monthly; the Study Group does not have an exact frequency in mind.

Separate search options for LCSH and MeSH. As said in section IV, the Study Group recommends that Acorn provide the public with separate search options for LCSH and MeSH headings, each with only the one appropriate cross reference structure built into it. This would prevent the LCSH and MeSH cross reference structures from coming into conflict and making navigation confusing for the user. We are open to the idea of having a combined subject search, preferably labeled to help indicate that it combines subject heading systems, available as well. It would include both LCSH and MeSH headings along with all subject headings which are neither. Alternatively, we could include the latter subject headings only in the keyword index. We do believe that decisions concerning the OPAC require input from and consideration by public services staff.

Blind reference problem. The Study Group urges that Vanderbilt, through the Sirsi enhancement request process and any other means available, continue to lobby Sirsi to fix the blind reference problem in Unicorn. In the meantime, we recommend that the "error" message currently displayed to users be made more helpful, if at all possible. The Study Group strongly recommends against jerry-rigging the coding of authority records to "fix" this problem locally.

Deletion of "local" subject authority file. The Study Group came to the decision some time ago that this holdover from NOTIS-to-Sirsi conversion was more trouble than it was worth, and recommended that it be deleted. That has already been accomplished.

Use Marcive. The Study Group recommends that Marcive be used to provide authority records for ongoing cataloging, updates of existing authority records, and a basic MeSH authority file. Marcive can provide all of the services we wish to contract, at a competitive price. Library Technology staff already work with Marcive to load records they supply for government document records, and have found Marcive staff supportive.

Members of the Study Group also considered LTI's and WLN's services. We do not like LTI's creation of "homemade" authority records to supplement LC authorities, and feel that such an approach to authority control would not befit a NACO library. We did not find anything in WLN's services that made us favor them over Marcive, and we found their staff disturbingly unresponsive to requests for information.

The startup and ongoing expenses of Marcive's authority control services have already been presented to Library Administration and did not seem to be prohibitive. Their estimate for starting the update service for the existing authority file was between $1500 and $4000. The cost of authority control for ongoing cataloging, done in conjunction with the update service, is 9 cents per bibliographic record (this covers all name, LCSH and MeSH headings), so this cost would fluctuate based on cataloging activity. After the startup costs, we would be charged around $1500 per year, plus the 9 cents per bibliographic record. We will be charged 5 cents per record for authority records we add to or delete from our update service (this would not include authority records automatically added to the service as a result of Marcive's processing of our new cataloging); we expect this will not add up to much. If the estimates are correct, we can probably fund the ongoing work with salary money currently being used by students to bring authority records in one by one from lists of newly cataloged titles. Furthermore, currently, we export authority records from OCLC using the "loadauthimp" process, at a price of 12 cents or more per authority record, or get them from the Library of Congress through Smartport, which is time-consuming and somewhat unreliable. Authority processing by a vendor would eliminate many of these money and time costs.

The Study Group recommends that we purchase paper reports with the authority records. This will facilitate manual review by authorities staff and assist in the planning of tapeloads.

Series authority records excluded from updating. The Study Group recommends that a vendor not be engaged to provide updates for our series authority records. We record local analysis, tracing and classification practice extensively on our series authority records, and automatic updating of them would override the local information.

No "whole-database" processing. For reasons elaborated upon in section V, the Study Group does not recommend having a vendor process our entire current bibliographic database to identify relevant authority records and make bibliographic corrections.

No changes to bibliographic records. We do not recommend having a vendor make changes to bibliographic records, either backfile or new cataloging, at this time, for reasons given in section V.

MeSH authority file. We recommend option 3, above, for the creation of an MeSH authority file. We would obtain MeSH headings for all existing Biomedical bibliographic records, except those for periodicals and Historical Collections. This option is relatively inexpensive.