The University of Tennessee Libraries’ Name Authority Challenge

At the beginning of 2007, the University of Tennessee Libraries (UTL) transferred the creation of descriptive metadata for digitized manuscripts from the Digital Library Center (DLC) to the Technical Services Department. After archives were scanned and digitized in the DLC, digital surrogates of the manuscripts were created using the Text Encoding Initiative (TEI) schema. TEI is a markup language for representing structural and conceptual features of texts. It is used primarily for the encoding of documents in the humanities and social sciences and, in particular, in the representation of primary source materials for research and analysis. Files in TEI were sent to the cataloging department to be transformed into rich descriptive metadata using the Metadata Object Description Schema (MODS), UTL’s selected schema for digitized manuscripts.¹

As a requirement of using MODS, catalogers have to use controlled vocabularies to assign access points to the records. Soon after receiving their first batch of TEI encoded records, catalogers encountered serious difficulties in assigning personal names to the access points of MODS records. The following were the main problems:

• Difficulty in finding names in TEI records in the LCAF. This problem occurred because either the record did not exist in the LCAF or because names in the TEI records could not be matched with names in the LCAF because of the lack of sufficient biographical information in the TEI records to identify individuals. For example, proving that the individual in the TEI record was the same one listed in the LCAF was difficult because no data other than name were given.
• Inconsistency in the establishment of names not found in the LCAF. When names were not found in the LCAF, different catalogers assigned different headings for the same person, depending on the form of the name given in the manuscript. Entering the same individual’s name in many different ways can create a serious problem for future discovery and access.
• Difficulty in differentiating individuals with similar names within the same collection. Many people whose names appeared in the manuscripts in the UTL collection shared the same or similar names with relatives mentioned in the collections. Distinguishing between two or more individuals with similar names became difficult because little, if any, biographical data were provided in the manuscripts. To make matters worse, individuals sometimes were called only by nicknames or had very commonly used names, which were difficult to differentiate from other similar headings. These factors created confusion for catalogers and made the process of differentiation almost impossible.
• Uncertainty about how to handle misspelled names and other typographical errors in the TEI transcriptions. Sometimes errors were made in transcribing names from the digitized image to the TEI files. Catalogers did not know whether to go back and fix the misspelling by editing the TEI record or to create an access point using the form found in the manuscript, even if it were a misspelled form. Different decisions made by various catalogers brought more inconsistency to the access points.

The problems in assigning personal headings to the access points of MODS records demanded an effective method to handle name authority control in the UTL’s digitized collections.

Literature Review

Libraries and other institutions seeking to support name authority control in XML metadata have tried various approaches. Some have proven more successful than others, but none has consistently been implemented for XML documents. Some commonly mentioned approaches in the library literature include Metadata Authority Description Schema (MADS), MARC Extensible Markup Language (MARCXML), Encoded Archival Context (EAC), OCLC Linked Authority File (OCLC LAF), and the Automated Name Authority Control (ANAC). Most of these authority initiatives for non–MARC metadata are designed to handle authority control at the local level; only a few try to do so at the national level. Some are XML schemas for authority elements created for use in conjunction with particular XML bibliographic schemas. Others are conversion tools that convert MARC into XML records. Some authority initiatives claim to be automated, but usually these are really semiautomated approaches that apply a mixture of manual and automated approaches to generate authority control.

Designing a Practical Approach at UTL

After reviewing the library literature and analyzing the advantages and disadvantages of the different approaches to support name authority control in XML metadata, UTL decided that none of these approaches was appropriate for the local situation. Because of time and funding constraints and lack of technological support, UTL decided to design a different approach that would be customized for UTL. Several points needed consideration. First, the new authority control method had to be completely sustainable by the UTL catalogers. Sustainable in this context meant the method had to be cost effective and use a level of technology with which the catalogers were comfortable. Because the cataloging department could not hire a local programmer, they ruled out building a local XML name repository and decided to capitalize on existing staff knowledge instead. Second, the authority control had to be achieved within a reasonable amount of time. Because this work is time consuming, priorities for which names to establish and which ones to leave out had to be set from the beginning. These priorities will be referred to from now on as “establishment criteria.” The reason for using establishment criteria is that searching, verifying, and establishing each name found in the TEI records would be impossible because they number in the thousands. The criteria would determine the cases in which names would be searched and established in the LCAF. Third, the new authority method had to support a level of quality if UTL wanted to keep materials by the same creators together. Given these considerations, UTL decided to use a manual approach that keeps taking advantage of the largest name authority file available—the LCAF.

The method UTL implemented integrated all the previous points. The process is described in detail in the following section. Briefly, authority control is performed as soon as the DLC sends the TEI transcriptions to the cataloging department. Authority control, then, is performed before records are cataloged, and only by one person to avoid future inconsistencies in establishing names. The person chosen to perform this task is one of the catalogers, who had previous experience creating authority records through the Name Authority Cooperative Project (NACO). After authority work is finished, headings are stored in a Microsoft Excel shareable spreadsheet. As soon as the spreadsheet is ready, catalogers are notified and TEI records are sent to them. The catalogers then have the necessary resources to catalog and create rich, descriptive MODS records with the least amount of effort.

The detailed process followed at UTL consists of the following steps: The librarian charged with authority control receives a batch of one hundred to three hundred TEI transcriptions with their digital images from the DLC. The files usually originate from many different collections in the university archives. The authority control librarian devotes approximately two weeks (full time) to authority work for this batch of records. After receiving the TEI files, the authority librarian opens and browses the files using XML Pad. First, she groups the files by collection name. She identifies the collection to which a TEI file belongs by looking at the tag <collection> under the <sourceDesc> in the TEI file or by looking at the second portion of the TEI identification number. In the example 0012_000060_000337_0000, “60” identifies the collection to which the TEI belongs. This is illustrated in figure 1. (The zeros serve as fill characters and are omitted by the authority librarian when noting these data in working files.) By grouping TEI files by collection, catalogers will start working with all TEI files in one collection before moving to the next one. The logic behind this approach is to become familiar with the finding aid of a single collection by reading it once instead of having to read it many times. Another reason is that names in TEI files within one collection usually relate to each other, making keeping track of family and other types of relationships easier.

Within each TEI file, the authority librarian browses the following sections to search for names:

• Title section, to retrieve the names of senders and recipients.
• Body section, to find names that “pop out” as important access points. If summary sections are provided, the librarian should browse through them as well.
• Signature section, to determine the preferred form of heading for the sender.

Names found in these sections become crucial because they will form access points for the MODS record, the equivalent of the fields 1xx (main entry fields), 6xx (subject access), and 7xx (added and linking entry fields) in MARC records. The authority librarian should pay attention to variant forms of headings as well.

The authority librarian makes a list of the names found, as well as of their variant forms, along with the record number of the TEI where the names were found. By recording this number, the librarian can later retrieve the exact location of these names in case more information is needed. In going through the rest of the TEI files, the authority librarian may encounter the same names, as well as other new names or variant forms of names, and keeps a list. This process is illustrated in figure 2.

After browsing the TEI files in one collection, the authority librarian looks at the names gathered so far and compares them to see if some names have been mentioned more than once using the same form or a variant one. She also counts the number of times each name is mentioned in different TEI files.

The establishment criteria are then applied to names in that particular collection. These criteria help determine which headings will be searched, verified, or established, and which ones will not. UTL developed establishment criteria that worked well in most situations. Names mentioned in at least three separate TEI files are searched in the LCAF and established if not found. The same process applies to names mentioned in the “Title” section of the TEI files as senders or recipients and to names that have a collection with their name (this can be checked in the “Collection” section of the TEI file). The one exception to the establishment criteria is the handling of names for prominent historical individuals. Because they are likely to appear in the LCAF, they are also searched. Figure 3 illustrates application of the establishment criteria.

For names that will be established according to the criteria, the authority librarian returns to the TEI files in which they were found. This is done by using the TEI record number that was noted on the list. When retrieving the TEI records, the librarian browses the text around the area where the names were found to get as much information—stated directly or indirectly—about the person as possible. Examples of useful areas to browse in TEI files are the “Title” section, which gives the date and place a letter was sent, and the “Body” section, which may provide information on people’s roles, relationships, and so on. The authority librarian annotates this information, along with the variant forms of the name found. The result might look like this:

Jacob Breck, Jab Breck, Jacob B.; sender of letter from Franklin, TN in 1864 to Philadelphia, Penn; judge; wife Lizzie.

These brief factual data will provide a general idea of who this individual was and when he or she lived. The data found can be expanded later through further research in outside sources.

After all biographical facts available in TEI files have been annotated, the authority librarian then consults various research tools such as finding aids. The University Archives, which own the original texts for the TEI files, have created finding aids, many of which are online. Tennessee state and county archives also may contain related finding aids. These tools may provide information on the person’s time period, family, place of residence, and more details. This information will be used by the authority librarian to place this person in context, see with whom he or she associated, and differentiate the individual from others with similar names when searching the LCAF.

If finding aids do not provide enough information about a person or are not available, the librarian searches other outside sources such as Google Book Search. This tool provides the ability to search sections within long texts of reliable resources that are freely accessible online. Other useful and freely accessible websites for historical biographical research include the Political Graveyard—A Database of Historic Cemeteries (http://politicalgraveyard.com), the state finding aids via the state library or state historical society websites, Genealogybuff.com, the Biographical Directory of the United States Congress (http://bioguide.congress.gov/biosearch/biosearch.asp), and the Civil War Rosters website (www.geocities.com/Area51/Lair/3680/cw/cw.html). The latter can be searched by soldier, regiment, and more. In addition, the authority librarian searches the Tennessee Genealogy and History Web (TnGenWeb.org)—other states may have similar sources. Other fee-for-service genealogy databases are also available. When searching historical personal names online, a useful tip for best results is to search using very specific factual data. For example, if the only information available from a TEI transcription about “Lord Cornwallis” indicates that he was alive during 1791 and wrote from Blount County, these facts should be integrated into the search.

As relationships between individuals start becoming clearer, the authority librarian should illustrate the relationships using visual aids in addition to noting the information. Visual aids such as genealogical trees, arrows, and diagrams can prove useful to represent relationships between individuals. Visual aids are important for the authority librarian because she may need to consult these aids to create names, and they are important for the rest of the metadata team, who will assign the names as access points in the MODS records. Understanding relationships among the individuals in the digitized transcriptions is crucial to create useful access points for the records.

After gathering enough data about a particular individual, the librarian searches the name in the LCAF. At this point, she will have enough biographical information to distinguish that individual from others with similar names in the LCAF. If the heading is found in the LCAF, the authority librarian copies and pastes the heading into a local Excel spreadsheet, along with its cross-references and notes. The spreadsheet serves two purposes: It builds a local database of the established names found in UTL’s digitized archives and provides catalogers with a narrower list of established names that appear in the TEI files they will catalog, saving them the time and effort of searching the LCAF. Sometimes, the authority librarian has located extra information not already mentioned in the LCAF about the individual listed. This extra information, such as biographical details or other variant forms of names, can be added optionally to the LCAF record in order to enhance it. This additional information can help differentiate this person from others with similar names in the future.

If the heading is not found in the LCAF, the librarian searches the OCLC Connexion Bibliographic File for records that used this name in any of their access points. To search in these areas of the bibliographic files, the authority librarian performs keyword searches using the index labels “au” (author) and “su” (subject). Searching in the OCLC Bibliographic File is a step required before establishing any heading in the LCAF. This search often leads to records that mention variant forms of this person’s name as well as extra facts not discovered previously.

After searching the OCLC Bibliographic File, the authority librarian establishes headings that were not found in the LCAF using the biographical information gathered to this point. Headings can be established locally or nationally, depending on the institution’s involvement with NACO. Libraries that are NACO members or part of a NACO Funnel Project have the option of making name contributions nationally. A NACO funnel project is a group of libraries who together are authorized to contribute name authority records to the LCAF. On the other hand, libraries that are not NACO members will not have the option of making national name contributions and will have to establish them locally. UTL has the option of making name contributions to the LCAF because it is a member of the Tennessee NACO Funnel Project. When establishing a heading, the authority librarian includes all the cross-references and factual data found previously in the research that may prove useful for the future. After establishing a heading in the LCAF, the authority librarian copies and pastes the heading into the local Excel spreadsheet with all the other LCAF names already found in OCLC.

After the chosen headings from one particular collection have been searched and established, the authority librarian browses the TEI files of the next collection, repeating the steps described above until all collections in the batch of TEI files are completed.

After names that met the establishment criteria have been searched or established, the lists of names that did not meet the establishment criteria remain. These lists are kept by the authority librarian in case any of the names need to be established in the future. Each list contains the TEI record numbers indicating where those names were found and can help retrieve the records if they are needed later.

After the authority librarian completes these steps, the authority work is considered completed. The tools and resources needed for cataloging metadata are then placed in a shareable department server. These include the digitized files in JPEG, transcription files in TEI, visual tools, and the Excel spreadsheet with the authorized name headings. The catalogers are then prepared to start creating MODS descriptive metadata with the least amount of inconvenience.

Implementing this authority control process before the rest of the metadata production starts solved the problems UTL initially faced when trying to assign name access points to MODS records without authority control. This approach to authority control solved both the difficulty in finding TEI names in the LCAF and the inconsistency in establishing names if they were not found there. Now that the authority librarian provides all the necessary authority work, the catalogers will not have to worry about searching these names in the LCAF or establishing them. The catalogers will find the established forms plus their variants in a local, shared Excel list.

Placing authority control before the rest of the metadata process permitted the catalogers to focus on the rest of the description. It solved the difficulty of differentiating individuals with very similar names within a collection by providing useful biographical information. The use of qualifiers and other attributes in authority control also helped in this purpose. The provision of visual aids such as genealogical tables helped catalogers throughout the process of visualizing family relationships and helped to diminish confusion about similar names.

The problem of misspelled names and other typographical errors that occurred when transcribing names from the original text to the TEI files was also solved with this authority method. By receiving the TEI files with their digitized images as a first step, the authority librarian had the opportunity to catch transcription mistakes and fix them before the catalogers had the chance to discover them.

Conclusion

As evidenced throughout this paper, many libraries and institutions are looking for ways to turn necessary tasks over to machines, but experience suggests this is not yet possible for name authority control in XML metadata. The efforts created so far to achieve this goal, besides being costly and work intensive, have proved to be ineffective and unreliable. Most do not address the issue of how to extract or harvest names directly from the XML records and transform them into useful access points, but focus on how to encode the access points into XML authority schema. The few endeavors that have tried to harvest names directly from XML records have proved not to be completely reliable in their processes of matching and linking names, making them dependent on human effort.

In addition to not addressing how to select and extract access points from the XML records, most of these endeavors require labor-intensive encoding of authority data into XML schemas and, subsequently, the creation of a local XML name repository to store and manage these records. Building an XML name repository is a task that requires a high level of technological background most catalogers lack. For this reason a programmer will have to be hired to build a name repository, and this is an expensive approach not many libraries can pursue. Furthermore, creating authority data to be stored in a local repository will only benefit the local institution, causing inconsistencies and duplication of efforts between different institutions that try to set up access points for the same individuals. Initiatives that tried to avoid the duplication of efforts in name authority control—by creating a national XML name repository to share authority data and make it interoperable between different institutions—have not been successful because the XML authority data needs to be shareable to be interoperable between the national repository and the other institutions. To date, this has not been successfully achieved.

In contrast to these approaches, UTL’s method to support name authority control in XML metadata is effective, reliable, and cost effective. It addresses the issue of extracting names directly from the XML documents and turning them into useful access points that can be shared nationally through the LCAF, thus avoiding duplication of efforts and benefiting all libraries who may share the same access points.

UTL’s approach is simple and can be used by other libraries and institutions that face similar issues when trying to support name authority control in their XML metadata. Common problems such as inconsistency in the establishment of names, difficulty in differentiating individuals, and deciding which names to turn into access points can be solved by implementing this method before creating any descriptive metadata for digitized transcriptions. Regardless of the local XML schema used, this approach can be applied in the same way to different collections.