- Digitizing the Collection
- Cataloging the Collection
- Interoperability between the Library of Congress and Brown University
The Brown University Library received an award in the 1996/97 round of the Library of Congress/Ameritech National Digital Library Competition to support the digitization of this collection of sheet music. Bibliographic records and image files were delivered to the Library of Congress for indexing as part of American Memory. Records in the Brown University Library catalog includes links to the digital reproductions mounted at the Library of Congress.
Project staff, along with staff from the Brown University Library Systems Office, staff from Brown's Scholarly Technology Group, and in consultation with the scanning vendor, ImageMakers, worked closely with staff at the Library of Congress to determine appropriate file formats and image sizes for the digitized sheet music. Since access was to be supported from image files stored at the Library of Congress, conventions for naming the files were also determined in consultation.
Choosing Image Specifications
Image sizes were determined after experimentation with several different sizes and formats. Goals included:
- a relatively small "page-turner" image that would provide sufficient detail to be identifiable without requiring excessive load time;
- to a "printing" version that would satisfy most of the needs of performers, musicologists, and cultural historians, as well as the general public;
- a higher-resolution "download" version that would permit close-up inspection of an item.
- an archival master for retention at Brown University Library.
Three "worst case scenario" images were chosen for experimentation: a very elaborate and detailed multicolor cover; a highly acidic black-and-white notation page (very browned); and a very faint blue-printed back cover with notation. ImageMakers digitized several versions of each, at different resolutions, and mounted them on a trial web site.
Over a period of several weeks, staff inspected and compared the images, viewing them on a variety of monitors attached to computers of different speeds. It became apparent that, for legibility and simple identification, a 400 pixel tall page turner was most appropriate. It was further demonstrated that scanning at 300 dpi made a significant difference in detail and color fidelity, particularly in the crucial printing version, the "workhorse" of the project. Further, the 860 pixel size of the printing version allows the image to print on one page, as tested on a variety of printers.
|Sizes and digital formats for images|
|Master Image:||TIFF, 300 dpi, LZW compressed
(retained at Brown University Library)
|High-Resolution Download Image:||JPEG, 1650 pixels tall|
|Printing Image:||JPEG, 860 pixels tall|
|Page-Turner Image:||JPEG, 400 pixels tall|
The images were produced at a tonal depth of 8 bits per pixel for grayscale (usually interior notation pages for works post-1890, and full works 1850-1890) and 24 bits per pixel for color (usually front covers and occasional back covers for post-1890 works). JPEG compression was approximately 10:1.
Capture, Quality Review, and Delivery
Digital images were produced by the vendor, ImageMakers, of Providence, Rhode Island, on a Mirage D-16L flatbed scanner. The image files were delivered on CD-ROMs to Brown University Library where they were inspected for quality and fidelity to naming conventions. Quality assurance was performed on a Power Macintosh 7300/200, running System 7.5.3, with 64 MB of RAM, a 2 GB hard disk and an Apple Vision 1710 17" color monitor, and using Adobe Photoshop 4.0 and Graphic Converter 3.0.2 software. To facilitate project management, a database was constructed using FileMaker Pro. Bibliographic records from the Library's Innovative Interfaces online catalog were downloaded into the project management database.
The online collection consists of 1,305 pieces of sheet music. In all, 7,031 pages were scanned (an average of 5.4 pages per item). The 21,093 digitized images, 3 for each page, were sent to the Library of Congress on CD-ROMs for mounting as part of American Memory.
At the Library of Congress, "page-turning" dataset files were generated for each piece of sheet music. The files were derived automatically from a list of item titles with corresponding identification numbers and the directories of image files named consistently. Each page-turning file incorporates the title of the piece and structural metadata to support a user interface that allows navigation through the sequence of pages and access to the larger images for each page. To allow presentation of covers of several close variants described in a single catalog record to be displayed simultaneously over a modem connection, smaller thumbnails were derived automatically for cover pages.
In 1980, Brown University Library was awarded a grant under Title II-C of the Higher Education Act to catalog 1,700 titles from the African-American Sheet Music Collection at the John Hay Library. This grant, awarded both because of the significance of the material to be cataloged and the innovative nature of the project, enabled the Library to input records to the RLIN scores database and to add numerous authority records for hitherto unestablished composers, lyricists, and performers.
Brown University Library staff members Sarah Shaw and Lauralee Shiere, in their Introduction to Sheet Music Cataloging and Processing: A Manual, (Music Library Association: 1984), based on their experiences working on the Brown project, note: "The project was designed to explore the full potential of the MARC scores format and to catalog the sheet music in a manner similar to rare book treatment. Forty-two data elements and seventeen access points were picked up in the cataloging records. This special collection treatment of the sheet music cataloging can be seen, for instance, in the detailed expression of paging given in the physical description area of the record and in the notes used to describe content of advertisements." The Manual was for many years the primary resource for sheet music catalogers nationally, and it is the cataloging from this project that has been used for the bibliographic records linked to the digital reproductions in the African-American Sheet Music collection.
Working in the Cataloging Module of the Library's Innovative Interfaces online catalog, project staff edited existing bibliographic records for each title selected to reflect the existence on an online resource, The edited records were "cloned" and a separate record for the online resource was created for each selected title. Multiple copies of a single title are reflected in a single bibliographic record for the original artifacts and another single bibliographic record for the online resources.
Many titles in the collection exist in multiple copies in which there are minor variations (usually in the performers whose portraits are included on the front cover or in back cover advertisements). Frequently, not all copies were digitized: it often occurred, for example, that copies 1,3,4,6,7, and 10 out of 12 copies were digitized. Selections were made on the basis of condition; a poor condition item with only minor variations was often not selected. In the case of exact duplicates, the item in better condition was selected. In most cases, copies for all different performers portrayed on the sheet music covers were digitized.
The records for the online digitzed resources were isolated in a review file and sent by FTP to the Library of Congress for incorporation into the American Memory site. In addition, the records were loaded in the OCLC and RLIN databases, assuring worldwide access to the images of the sheet music.
For this collection, the image files are mounted at the Library of Congress and presented through the same page-turning interface as used for many of the Library's own American Memory collections. Copies of MARC catalog records created at Brown are held in the online catalog at Brown, indexed using InQuery by the Library of Congress for American Memory, and incorporated into union catalogs at OCLC and the Research Libraries Group. All these records link to the digital reproductions mounted at the Library of Congress.
Presenting the Digital Reproductions
The digital reproduction for each piece of music consists of images in several sizes for each page and the related page-turning file. The page-turning file incorporates the "structural" metadata necessary to allow a user to navigate through a sequence of pages and select larger images to view if desired. The on-screen presentation of an item is generated dynamically from these components by a script in the Perl language.
Linking from Catalog Records to the Presentations
Since duplicates of the MARC records will exist in several different catalog systems, it is important that the links in these records to web presentations of the digital reproductions not need to be updated if the corresponding files are moved, for example, when computer systems are upgraded. To this end, the Library of Congress assigned persistent identifiers (Uniform Resource Names or URNs) using its Handle Server to the presentation of each piece of sheet music. In contrast to a Uniform Resource Locator (URL), a URN is independent of a particular storage location; if the Library of Congress needs to move files when it reorganizes or upgrades its storage systems, a change in the Handle Server will ensure that access to the items is possible through the same links even when the files are moved. There will be no need to change every copy of every catalog record that links to these items.
The handles have the general form loc.award/rpbaasm.NNNN where NNNN represents a unique project ID assigned by Brown to the particular item. The handle for item NNNN has been registered to "resolve" to the page-turning presentation script with NNNN specified as a query parameter. The Library of Congress runs a Proxy Handle Server (hdl.loc.gov) to allow resolution of handles with regular browsers. Through the proxy server, the handle above can be resolved through the URL http://hdl.loc.gov/loc.award/rpbaasm.NNNN
At Brown, URLs in this form were recorded in subfield $u of 856 fields of the catalog records in the university's online catalog. Copies of the records were both transmitted to the Library of Congress and also released for inclusion into OCLC's WorldCat database and the Research Libraries Group's union catalog.
Transformation and Indexing of Catalog Records
Brown University chose to generate catalog records for the digital reproductions that were distinct from the records for the original items. However, American Memory configurations for display and indexing are based on the single-record practice for one-for-one digital reproductions used by the Library of Congress. This practice, documented in Draft Interim Guidelines for Cataloging Electronic Resources, adds information about the digital reproduction to the record for the original item. The records delivered by Brown University were therefore modified automatically in minor ways to allow effective integration into American Memory and facilitate coherent searching across collections. Modifications included removing fields with information about the creation of the digital reproduction and adding local boilerplate fields to distinguish items in this digital collection from other American Memory collections.