The OSU Press Publications collection in the Knowledge Bank (http://hdl.handle.net/1811/131) contains open-access monographs no longer in print and new monographs for which the full-text is embargoed five years following date of publication. The OSU Press staff continue to add the new embargoed titles as they are published using a DSpace submission workflow mediated by SRI staff. The project to digitize and archive the open-access monographs began in August 2005. An SRI librarian created the core element set for the collection and designed the metadata workflow in consultation with the OSU Press community. The three hundred open-access titles contained in the collection were selected by OSU Press, digitized by a vendor funded by OSUL, and added to the collection by SRI staff and students. The open-access titles were digitized in three groups of one hundred and individually archived in the Knowledge Bank as the files were received from the vendor in 2006 and 2007. The monographs (metadata and PDF files) were added individually to the collection via the item submission user interface.

The metadata workflow for archiving the digital open-access monographs took advantage of the fact that the OSU Press print monographs were fully cataloged with MARC records in the library catalog. The workflow was designed to manually use the MARC catalog metadata to create DC item records in the Knowledge Bank. The core element set document listed the MARC data to be reused for DC fields including creator, date issued, description, description (table of contents), title, title (alternative), identifiers, publisher, relation (is part of series), and subject (Library of Congress Subject Headings (LCSH)). The SRI staff and students who submitted the items to the Knowledge Bank used the title, author, date, and any other pertinent information found in the PDF file of the digital monograph to search for and identify the correct MARC record in the catalog for the matching print version of the monograph. The staff member or student copied the reusable data from the catalog record and pasted it into the appropriate fields in the input form as they entered metadata and uploaded the PDF for each title. Using the library catalog metadata this way significantly reduced the amount of keying needed to input the metadata.

While manually copying and pasting the catalog metadata for the repository items saved keying, it was still a labor-intensive process. To improve the workflow efficiency, the author developed a new procedure in September 2007 to automate the process of reusing the catalog metadata for the Knowledge Bank records following receipt of the final one hundred digitized monographs. The design of the new workflow began with the core element set document for the OSU Press collection. The author revised the document to serve as a descriptive metadata crosswalk for the transformation of the metadata from MARC to qualified DC. The one hundred MARC records for the transformation were extracted from OSU's Innovative Interfaces Millennium catalog. The author then used the MARC-to-MARCXML function in MarcEdit to transform the MARC output file to MARCXML.

The author used an existing MARCXML to DC XSLT stylesheet available from the Library of Congress as a template and modified it for the requirements of the project.¹⁵ The first modifications to the template changed the DC schema used by the stylesheet. The Library of Congress stylesheet uses oai_dc, the XML schema for the Open Archives Initiative Protocol for Metadata Harvesting (OAI-PMH) and outputs the MARCXML metadata in the OAI format of DC. This OAI Simple DC schema was replaced with the qualified DC schema required for importing items in batch to DSpace. Figure 1 shows an example of each format. The author registered, or added, the modified stylesheet to the local application of the MarcEdit software for testing. An XML conversion function was added to the MarcEdit tool to use the modified stylesheet added to the software. The author then used the OSU Press MARCXML file as the test input file for the transformation. The new stylesheet successfully transformed the MARCXML file and output a file of Qualified DC in the DSpace format.

Following the conversion of the stylesheet to the DSpace Qualified DC schema, the author used the metadata crosswalk for the project as a guide to modify the MARC data fields in the stylesheet. The MARC fields not used in the OSU Press profile were removed from the stylesheet, including the 506 (restrictions on access), 521 (target audience), 530 (additional physical form available), 653 (index term—uncontrolled), 662 (subject added entry—hierarchical place name), and 856 (electronic location/access) fields. The various 76x, 77x, and 78x relation fields also were removed. The MARC fields needed for the transformation not present in the stylesheet were added, such as the MARC 440 and 830 series fields. The MARC table of contents field 505 was removed from the MARC 5xx fields for the DC description field and added as a separate DC table of contents description field. A DC contributor field was added and the MARC fields that were mapped to DC creator were split between creator and contributor. The MARC 6xx subject fields remaining in the modified stylesheet were restricted to LCSH by specifying a second indicator of “0” (zero) for each MARC 6xx data field. The MARC subfield codes also were modified on the basis of the requirements of the crosswalk. For example, the subdivisions “v” (form), “x” (general), “y” (chronological), and “z” (geographical) were added to the existing subfield codes for the MARC 6xx subject data fields. The author then added the revised stylesheet to the local MarcEdit software application. The MarcEdit XML conversion function for the project, modified to use the revised stylesheet, transformed the MARCXML file to Qualified DC. The result of the transformation was one XML file with all of the DC records for the OSU Press items. The next step in the new workflow involved preparing the DC metadata and the PDF files for batch loading into the Knowledge Bank.

The OSU Press stylesheet still had limitations that needed to be addressed. The manual cleanup of the XML file included removing excess white space, unwanted punctuation, and the namespace declarations occurring within each DC record. Double-dash delimiters were also added between the subject subfields. After the author corrected the metadata, the OIT systems developer wrote a short Perl script that separated the DC records contained in the one XML file into individual dublin_core.xml files and automatically creating the entire archive directory (PDF files and DC metadata). The author had manually pasted the PDF file names to the end of each record in the XML file during the metadata cleanup to facilitate this parsing. The OIT systems developer successfully batch loaded the completed archive directory (DC metadata and PDF content files) into the staging, or test, instance of the Knowledge Bank.

The successful batch could have been loaded into the production instance of the Knowledge Bank, but this new OSU Press workflow became a proof-of-concept exercise because the project's last group of one hundred monographs was manually entered into the Knowledge Bank for staff training purposes. Valuable experience was gained in the process of developing the automated workflow. The most important outcomes were the addition of a new automated workflow to repurpose library catalog metadata and a new batch loading process available for projects where the source metadata are contained in DC XML rather than in an Excel spreadsheet or a CSV file. As a result of the exercise, the author identified limitations of the new MARCXML-to-DSpace Qualified DC stylesheet that required further investigation to minimize the manual cleanup needed for the transformed metadata. The author refined the new stylesheet for use in a subsequent project for the OSU Byrd Polar Research Center Archival Program described below.

The OSU Byrd Polar Research Center Archival Program is a collaborative effort of the Byrd Polar Research Center (BPRC) and the OSU Archives. The BPRC, recognized internationally as a leader in polar and alpine research, is named in honor of the polar explorer Admiral Richard E. Byrd. Part of OSUL's Special Collections, the BPRC Archival Program collects, preserves, and provides access to historical polar-region material. The BPRC community in the Knowledge Bank currently includes collections for reports, publications, conferences, photo albums, and two oral history collections: the Polar Oral History Program collection (http://hdl.handle.net/1811/6039) and the Antarctic Deep Freeze Oral History Project collection (http://hdl.handle.net/1811/35321). The Polar Oral History Program collection archives the program's transcripts of interviews documenting the early years of American polar exploration. The Antarctic Deep Freeze Oral History Project collection archives the transcripts of interviews with pioneers of South Pole exploration documenting United States involvement in Antarctica during the 1950s.

The Polar Oral History Program collection currently contains eighty-five transcript items. The first item was submitted to the Knowledge Bank in 2005. The metadata workflow for the collection was designed by a former metadata librarian in consultation with the community representative, the polar curator. The metadata librarian created the core element set document and a specialized input form for the collection. Following the workflow set up by the former SRI metadata librarian, the polar curator enters the items individually using the input form customized for the collection in groups of approximately five to fifteen interview transcripts. Similar to the availability of MARC records for the print versions of the digital OSU Press monographs, MARC records for the sound recordings used to transcribe the interviews are available in the library catalog. Approximately 80 percent of 104 interview audio tapes were represented in the catalog at the beginning of the Knowledge Bank project.

The core element set for the collection reused the available catalog metadata for the DC title, creator, contributor, date created, relation (is part of series), subject (LCSH), and subject (other) fields. The former metadata librarian added the values for the creator, contributor, subject (LCSH), and subject (other) fields to dropdown menus in the input form. The DC fields that have constant (default) metadata for the collection, including publisher, rights, relation (is part of series), sponsorship, and type were added to the collection item template used to prepopulate fields in the input form.

Although labor-intensive, the work done at the beginning of the project to create the customized input form significantly reduced the keystrokes required of the polar curator to submit each item. Because not all of the audio tapes were cataloged when the input form was created, the drawback to this workflow was the ongoing maintenance. During the last several years, new completed groups of transcripts included some that did not have metadata in the input form dropdown menus, either because the matching audio tapes were not yet cataloged or were cataloged after the original input form was created. The author added names and subjects to the dropdown menus as MARC records were added to the catalog by OSUL's Special Collections Cataloging (SCC).

The polar curator requested a new BPRC oral history collection for thirty-nine Antarctic Deep Freeze Oral History Project transcripts in October 2008. The author, in assessing the metadata requirements of the project, had several options. The existing metadata workflow for the Polar Oral History Program could have been extended to include the new collection. To use the same workflow for the new collection, the author would need to manually copy and paste all of the reusable catalog metadata to the XML input form (either the same input form as used by the Polar Oral History program or a new input form modeled on it). Another option would have been for the Polar Curator to use an input form that had default data prepopulated via the item template, but did not have controlled vocabulary dropdown menus for the reusable, variable, catalog metadata. This second option would have eliminated the labor-intensive work on the part of a metadata librarian but would have shifted the rekeying to the polar curator who would copy and paste from the catalog records to input the items. Given the amount of redundant work and rekeying involved, the first and second options were not ideal. The author, in consultation with the polar curator, decided to try a third option. The new automated workflow tested with the OSU Press open-access monographs collection was modified and used for both oral history collections.

For the new workflow design process the author first addressed the existing Polar Oral History collection while SCC cataloged the Antarctic Deep Freeze audio tapes for the library catalog. The author used the XSLT stylesheet created for the OSU Press project as a template and modified it for the Polar Oral History collection. The fields for the new stylesheet were based on the Polar Oral History core element set. The constant data fields (default metadata) for the Polar Oral History collection, including publisher, rights, relation (is part of series), sponsorship, and type were added to the stylesheet as text elements, which serve to output the text enclosed within the element.

The author addressed many of the limitations of the OSU Press stylesheet in modifying the XSLT for the Polar Oral History Program. Functions to normalize space and remove unwanted punctuation were added. A template was added so that the XML namespace declaration would not occur in each DC record. Figure 2 illustrates the additions the author made to the “subfieldSelect” template to allow for alternate delimiters. An example of the modifications to the Polar Oral History stylesheet to create double-dash delimiters in subject fields is shown in the 650 data field in figure 3.

With no new Polar Oral History transcripts to batch load, the author tested the Polar Oral History stylesheet with transcripts that were already in the Knowledge Bank. Following the same process as with the OSU Press workflow, a list of catalog records was made and MARC records were output. The Polar Oral History stylesheet and the Library of Congress utility stylesheet modified for alternate delimiters were added to the local application of MarcEdit, a new XML conversion function for the project was created, and the MARC OUT file from the catalog was used for the transformation. In contrast to the OSU Press workflow, the author did not generate a separate MARCXML file as an interim step. Instead, the MarcEdit option of beginning with a MARC file was chosen. With this option, the MarcEdit software does the prerequisite MARC to MARCXML conversion as part of the transformation to DC.

The amount of cleanup needed for the Polar Oral History DC metadata was significantly reduced from the author's previous attempt with the OSU Press stylesheet. The author, however, did need to manually remove the duplicate identifier fields from the DC XML file resulting from the duplicate identical MARC 099 local call number fields in the MARC records. Editing of the file was also required because the Polar Oral History core element set calls for variable metadata values that are contained in the MARC record but cannot be directly transformed from the MARC data. One example of this is the DC relation (is format of) field, which is used to indicate that the described resource is the same intellectual content of the referenced resource, but presented in another format. The core element set uses the relation (is format of) field for a statement about the audio tapes from which the transcripts are derived. The MARC 300 physical description field, which contains the number of sound recordings, was transformed into a DC description field. The relation (is format of) field was added to the stylesheet as a text element with a constant data statement and an underscore for the number of audio tapes. For each record in the DC XML file after the transformation, the author copied the number of audio tapes from the description field, added the number to the relation (is format of) field in place of the underscore, and deleted the description field. In the final editing step for the DC XML file, the author added the PDF file names for the transcripts to the end of each DC record in preparation for batch loading.

The Polar Oral History DC and PDF files were prepared for batch loading using the same Perl script written for the OSU Press workflow. The OIT systems developer used the Perl script to create the archive directory, and the collection was successfully batch loaded into the staging instance of the Knowledge Bank. The author plans to use the new Polar Oral History Program workflow for the next set of available transcripts.

The author used the Polar Oral History XSLT stylesheet as a template for the Antarctic Deep Freeze Oral History Project workflow and created a core element set for the new collection in consultation with the polar curator. The core element set was used to determine the fields needed in the stylesheet. As with the Polar Oral History stylesheet, constant data fields for the Antarctic Deep Freeze collection were added as text elements to the stylesheet. The Antarctic Deep Freeze stylesheet also uses the same modified Library of Congress utility stylesheet used for the Polar Oral History workflow. The XSLT stylesheet used for the Antarctic Deep Freeze project is available in appendix A. When the cataloging for the Antarctic sound recordings was completed by SCC, the author gathered the MARC records in a review file and output them to a personal computer. The MARC OUT file was transformed to qualified DC utilizing the Antarctic Deep Freeze stylesheet added to the local application of the MarcEdit software. Appendix B includes an example comparison of Antarctic Deep Freeze metadata in the original MARC format and in DSpace qualified DC format for batch loading.

The DC XML file for the Antarctic Deep Freeze project required cleanup similar to that of the Polar Oral History records. For example, the author modified the relation (is format of) fields and removed the duplicate identifier fields resulting from duplicate fields in the MARC records. As with the previously described projects, the PDF file names were added to the end of each record, and the Perl script written for the OSU Press project was reused to create the DSpace archive directory. The new collection of thirty-nine transcripts was successfully batch loaded into the staging instance of the Knowledge Bank. The Polar Curator reviewed and approved the collection on the staging server, and the collection was successfully batch loaded into the production instance of the Knowledge Bank on April 23, 2009. Figure 4 is an example of the public simple (short) item view for an Antarctic Deep Freeze transcript record in Knowledge Bank.

The automated workflow for the Antarctic Deep Freeze collection worked well. The XSLT transformation process went smoothly because of the previous work done to optimize it during the OSU Press and Polar Oral History tests. The author plans to further investigate XSLT options to reduce manual metadata cleanup for transformed records. However, the manual cleanup of the DC records could have been mitigated if different choices had been made regarding the desired Qualified DC metadata for the collection. Balancing metadata quality and completeness with efficiency is always an issue when creating metadata, and the same challenge is present when repurposing metadata. The author could have eliminated manual transformation procedures if the polar curator and author had decided to change the metadata profile. For example, the relation (is format of) field could have been a constant data field that stated audio tapes were available without providing the number.

The Perl script written by the OIT systems developer for the initial OSU Press test proved a successful way to create the batch loading archive directory for DC XML file source metadata. However, the author plans to further investigate an automated way to add the PDF file names to each record in the DC XML file needed for the current Perl script to create the batch loading archive directory.