Child pages
  • Grant Information

IMLS Variations3 final performance report part 1 2009-12-29.docx

Variations3: An Integrated Digital Library and Learning System for the Music Community
Indiana University Digital Library Program
IMLS National Leadership Grant LG-05-05-0209-05
Project Director: Jon W. Dunn, Associate Director for Technology
 

Final Performance Report Narrative: October 1, 2005 – September 30, 2009
Submitted December 29, 2009

This report provides an overview of the activities and outcomes of the Variations3 project led by the Indiana University Digital Library Program and supported by a National Leadership Grant from the Institute of Museum and Library Services from October 1, 2005 through September 30, 2009.

Project Summary

The primary goal of the Variations3 project was to develop a digital music library and learning system that could be easily deployed by a wide range of college and university libraries to integrate access to online recordings and scores into teaching, learning, and research activities. In addition, the project sought to streamline the process by which metadata could be created to improve the experience of users searching for music. In March 2009, Indiana University released the Variations digital music library system as free open source software, available for download and use by other institutions and the software is currently in production use at six institutions including Indiana University, with several additional sites currently evaluating or testing the system.

Project Activities

There were four major areas of activity in the Variations3 project: software development, software testing, metadata research, and dissemination.

Software Development

The software system that was used as the starting point for development work in the Variations3 project was originally created by Indiana University (IU) between 2000 and 2005 as part of the Creating the Digital Music Library project, also known as Variations2, [1] which was supported by a Digital Libraries Initiative – Phase II (DLI2) grant from the National Science Foundation. The resulting system was put into production at the William and Gayle Cook Music Library at Indiana University, replacing an earlier digital audio delivery system known as VARIATIONS that had been in place since 1996.

The Variations software developed through the Variations2 and Variations3 projects enables institutions such as college and university libraries and music schools to digitize audio and score materials from their own collections and provide those materials to their students and faculty in an interactive online environment, while respecting intellectual property rights. Faculty and students can create bookmarks and playlists for use in studying or in preparing classroom presentations, allowing easy access later on to specific audio time points or segments, and can annotate sound recordings and musical scores via easy-to-use graphical tools. A key feature for libraries is a flexible access control and authentication system, which allows library staff to set up access rules based on their own local institutional policies.

The Variations3 grant proposal identified four major issues to be addressed in order to make the Variations2 digital music library system in production at IU usable by other institutions: 1) support for more common technology infrastructure; 2) improved system reliability and administration tools; 3) support for easy ingestion of content into the system and improved discovery of content through existing MARC records; and 4) access to subscription music services alongside local content through the same user interface.

Over the course of the four-year project, seven versions of the Variations system were released to test sites. Initial releases focused on the changes necessary to allow Variations to be installable at institutions other than IU, including making many previously hard-coded settings configurable, and adding support for the popular MySql open source relational database management system. Later releases included additional features, such as support for more authentication methods, the ability to import CD track metadata from other sources, a Web-based audio player and access manager, access to content from the third-party subscription streaming service DRAM Online, as well as many bug fixes to improve system reliability and stability.

Software Testing

The Variations3 project used a “test site” model for validating the viability of the Variations system at institutions beyond IU. The four Variations3 test sites were The Ohio State University, University of Maryland, New England Conservatory of Music, and the Tri-College Consortium consisting of Haverford, Swarthmore, and Bryn Mawr colleges. These sites were purposefully selected to represent a wide range of types of institutions and IT environments. Test sites were expected to install the Variations system locally, ingest audio and musical score content, and conduct pilot tests of the system using a selected user population, typically the faculty and students in one or more classes using online materials as part of their syllabus. Each site was provided with an Intel/Linux-based server on which to run the Variations system, as well as access to the Variations code and documentation. Initial visits were made by project staff to each site to discuss the system, its functionality, and its technical and support requirements with a variety of local stakeholders, and additional support was provided over the course of the project through e-mail and telephone contact, remote access by IU staff to test site servers, and in some cases, in-person visits by IU staff.

Pilot testing was originally expected to begin in the fall semester of 2006, approximately one year into the project, but did not start until spring 2007. The process of getting test sites up and running took longer than originally thought, due to a variety of issues. The most time consuming of these was the execution of Memoranda of Agreement between Indiana University and each of the test sites that passed muster with the appropriate administrators and legal authorities at each institution. In addition, availability of technical staff to participate in various steps in the process of selecting server hardware, installing the server operating system (Linux) and software prerequisites, and installing and configuring the Variations software was a problem at some institutions. However, all institutions were able to get the system up and running and conduct pilots involving faculty and student use. Feedback from test sites was obtained via e-mail, telephone conference calls, and visits by project staff to test sites, and all four test sites are still using the Variations system as of the completion of the grant project.

Metadata Research

The third major area of activity on the project was the enhancement of the musical work-based metadata model originally developed as part of the NSF-funded Variations2 project and the investigation of techniques for streamlining the process of creating such work-based metadata.

In particular, we proposed to investigate aligning the Variations2 model more closely with the model described in the International Federation of Library Associations (IFLA) report on Functional Requirements for Bibliographic Records, better known as FRBR. [2] In the course of this project, we conducted an analysis of what would be needed to align the Variations2 model with FRBR and reviewed outcomes from experiments in “batch loading” of MARC records (described below). The results of this work, combined with advice from the project’s advisory board, led us to conclude that this would be a useful activity. However, due to other development needs, we had less time for metadata activities on the project than originally anticipated, and our analysis suggested more time would be needed than was originally estimated, so focusing a new project on this activity was the right way forward (see “Next Steps” section at end for additional details).

In addition, we proposed to further streamline the process for creating records through the investigation of four approaches:

1. Maximize automated mapping of data from MARC Bibliographic and Authority records into Variations3 records. We conducted three iterations of “batch loading” experiments during the Variations3 project. The results from these were very promising and contributed to our conclusion that a FRBRized model populated with data that began as MARC/AACR2 is feasible. These experiments resulted in a much more robust Work identification and MARC mapping algorithm [3] that we made publicly available.

2. Implement a cooperative cataloging workflow. We developed specifications for a cooperative cataloging workflow but did not have development time to implement this feature.

3. Experiment with import of metadata from non-library sources. Project staff conducted experiments with the open MusicBrainz [4] service to investigate whether it could be a useful source of data for populating work-based metadata records.

4. Experiment with allowing end-users to enrich basic metadata with their own contributions. We did not have available development time to devote to this activity.

Dissemination

The Variations3 project team disseminated information about its activities through presentations over the course of the project to a variety of different professional organizations, including the Music Library Association, International Association of Music Libraries and Documentation Centers (IAML), International Conference on Music Information Retrieval (ISMIR), International Society for Knowledge Organization (ISKO), Digital Library Federation, EDUCAUSE, and the Coalition for Networked Information (CNI). Several of these presentations appeared in published conference proceedings, and in addition, an article about the project was published in the August 2006 issue of Communications of the ACM , and this article and the project were subsequently featured in Science magazine. [5]

In February 2008, IU released the Variations Audio Timeliner, a standalone version of the timeline audio annotation feature from the full Variations system, as free open source software for Windows and Mac OS X, under a BSD license. In February 2009, IU released the full Variations system, also under an open source BSD license. Both software packages are available for download from the Variations open source web site. [6] The release of Variations as open source software was widely publicized via e-mail list postings and a press release, [7] and was featured in numerous blogs and publications, including American Libraries and Campus Technology .

In early March 2009, project staff conducted a free online “webinar” explaining the features, benefits, and requirements of Variations, using the Adobe Acrobat Connect web conferencing system. The webinar was announced over various relevant e-mail lists. Over 90 individuals from over 80 institutions attended the webinar, and a recording is available at the Variations web site.

Project Audiences

The audiences for the open source release of the Variations digital music library system and most of the project’s publicity and dissemination activities are librarians, faculty, and staff in college and university libraries, instructional or academic technology units, and music departments who need to provide controlled online access to music audio and score content to their users, for use in teaching, learning, and research activities. The Variations Audio Timeliner is also directly installable and usable by music teachers, students, and others interested in music, without need for the full Variations digital music library system.

Analysis

The four goals identified in the original grant proposal were:

  1. To develop a turnkey digital music library system that integrates access to local and licensed music content and provides tools for use of music in research, teaching and learning.
  2. To implement and evaluate a work-based metadata model, as part of the system, that supports improved access to musical information for users.
  3. To implement this system at Indiana University and at four test sites of varying characteristics.
  4. To establish a model for distribution, support, and continued development of the system.

Project Achievements

As identified in points 1 and 3 above, a major overarching goal of product development was to create a version of Variations that would be implementable at institutions other than IU and would also be attractive to additional institutions as an open source solution. We have achieved this goal, with all four of the initial test sites adopting and continuing to use Variations. In addition, as of the writing of this report in December 2009, we know of over a dozen additional institutions evaluating, piloting, or planning to adopt Variations.

At the end of the project, we are in what could be characterized as a “community building” stage in the evolution of Variations. Support, promotion, and ongoing enhancement of Variations are still being provided by IU, with approximately .7 FTE devoted to these activities. This is not a long-term sustainable model and does not allow for sufficient development effort to continue enhancing Variations to add new features and functionality and adapt to continual changes in technologies and user expectations. However, our hope is that we can build the community of Variations users to a sufficient size such that it can be self-sustaining, with a critical mass of institutions contributing financial and/or technical resources toward the system’s ongoing development, in a “community source” model similar to that of the Sakai [8] collaboration and learning environment or the Kuali [9] family of university administrative systems. This is a challenging goal, given the current financial situation of many colleges and universities and the relatively specialized nature (i.e. the focus on music) of the Variations system. As one possible way of alleviating this problem, we are exploring the possibility of extending Variations to serve other multimedia delivery needs of libraries and their host institutions.

From the Variations3 project’s metadata work agenda, two significant achievements emerged. The first is a better understanding in the music and FRBR communities about how the FRBR model applies to music materials. The Variations3 project team created two white papers, which systematically reviewed the features of the FRBR and FRAD reports and compared them to the Variations2 metadata model and the discovery needs of music users. The second primary achievement of the project’s metadata activities is a plan for the use of metadata from third parties such as MusicBrainz. Experiments with searching recordings via the MusicBrainz API showed that for recordings that tend to fail work identification through MARC authority records, 76% are present in MusicBrainz with some structured metadata, providing automated work identification routines with additional data beyond what is available from the library community.

Obstacles and Lessons Learned

Despite the project’s achievements, we did encounter some obstacles. Some of these have been noted above, such as the delay in starting pilots at our test sites. However, we also underestimated the developer time required to make the necessary enhancements to Variations to make it fully usable at test sites and releasable as open source software that any institution could pick up and use. This led to less time spent on implementation of the metadata creation approaches investigated by the project.

In the course of the software development and evaluation activities on the project, we learned a great deal about the diversity of the IT and management structures supporting libraries and music programs in colleges and universities. Two observations stand out as particularly relevant to the success of Variations and other systems for management and delivery of digital and digitized content:

  • Educational institutions have a wide variety of interpretations regarding and practices surrounding copyright.
  • Colleges and universities have been slow to develop the infrastructure to acquire and support open source solutions, particularly the less wealthy schools. However there is a trend in the open source direction, a trend that could be aided by the existence of third-party hosting service organizations that could help less technically sophisticated or less well-staffed schools.

Metadata work on the Variations3 project provided us with some very hard lessons, each of which, however, can also be seen as an opportunity. The decision to move away from the Variations2 metadata model to a fully FRBRized one was difficult, but we believe most of our work towards the earlier model will be reusable in this new context, and the opportunity to standardize and participate more fully in the FRBR community is a welcome one.

The Variations3 project had initially planned to experiment with cooperative cataloging among the project’s test sites. We realized fairly early on that there was not much to be learned nor much economy of effort to be had by following the OCLC model of cooperative cataloging, where records must be manually downloaded and then re-uploaded to the central pool (assuming cataloging agencies are all good citizens) after changes are made. This model requires more manual work than should be necessary in a cooperative environment and does not ensure each institution always has the most recent updates to any given record. However, technical work to design a more efficient and more automated model proved to be a larger job than we had planned for. Therefore we downplayed the role of cataloging in general for the Variations3 project, and test sites did not choose to experiment very heavily with the project’s metadata model.

A third major lesson from metadata work on the Variations3 project is the need for a great deal of modularity in the Variations system. While music discovery systems are still ripe for exploration, implementing new systems and workflows specific to music discovery is perhaps too big of a barrier for many potential Variations implementers. During the course of the Variations3 project, an endeavor was begun to “split” use tools from discovery tools. This is a challenging process as the two functions in some ways are intertwined and some data is needed for both, but the project made some progress towards making this distinction a reality within the Variations system.

Project Impact

The following data help to show the impact of the Variations3 project:

  • At our informational webinar on the Variations open source release in March 2009, over 90 people attended representing over 80 separate institutions.
  • Through December 18, 2009, the variations.sourceforge.net web site showed the following download statistics:
    • Variations Audio Timeliner runtime: 1332 downloads
    • Variations Audio Timeliner source code: 101 downloads
    • Variations server runtime: 171 downloads
    • Variations source code: 130 downloads
    • Variations demonstration client runtime: 69 downloads

Other impacts are less quantifiable, but equally important. Many of the test sites and potential implementers we spoke with over the course of the project had previously implemented or were planning track-based “audio course reserves” systems rather than true “digital music library” systems with complete recordings and scores. While this is of course in part due to limitations in the current copyright and licensing regimes, we believe that the wider adoption and exposure of Variations achieved as part of this project will help to underscore in more institutions the need to think more systematically about online availability of library materials, as opposed to the more haphazard reserves model.

Perhaps the biggest impacts of metadata work in the Variations3 project are features that help to lay the groundwork for future change in library cataloging. We expect the renewed focus on FRBR and exploration of reuse of third party metadata in this project and its follow-on to have a significant impact on the library community. Currently the Variations system connects to multiple online providers in order to automatically provide track titles for CDs, and in follow-on work, we will be putting into practice additional reuse of metadata from non-library sources such as MusicBrainz. In light of ongoing calls for integrating metadata from other communities into library workflows, such as that from the Library of Congress Working Group on the Future of Bibliographic Control, [10] we believe these concrete demonstrations of this principle in action will help to move this type of activity forward in the library community.

What’s Next

As noted above, with the open source release of Variations and limited support provided by Indiana University, we are in the process of trying to build up the Variations community to a size such that it can be at least somewhat self-sufficient. However, additional development work is necessary to make Variations into a tool that is attractive to a very wide range of institutions via its ability to support the management and delivery of multimedia content beyond music. We are currently in discussions with various leaders in the library and instructional audio/video communities about potential new directions for Variations development, including opportunities for additional grant funding to support expansion of the Variations system’s capabilities.

On the metadata front, as our internal analysis and discussions with the project advisory board made it clear that moving Variations to a fully FRBRized data model was necessary for its sustainability and provided an opportunity to advance the state of the art in next-generation, FRBRized library catalogs, we designed a follow-on project to achieve both of these goals. That project, Variations/FRBR: Variations as a Testbed for the FRBR Conceptual Model, [11] was funded as an IMLS National Leadership grant from October 1, 2008 through September 30, 2011. The Variations3 project has also demonstrated a great need for technical architectures that support cooperative cataloging with truly shared records rather than independent copies. While in the immediate short term we do not have plans to implement a full cooperative cataloging model, we hope the systems and best practices that emerge from the Variations/FRBR project will provide an opportunity for more community work in this area.

Grant Products

A CD-R is attached to this report, containing copies of the products and publications resulting from the project, listed below. URLs are also provided for Web-based materials.

Software

Variations 5.0.7 and Variations Audio Timeliner 1.0.1 source code and runtimes. http://sourceforge.net/projects/variations/files/

Reports and Documentation

Variations Implementation Guide. http://wiki.dlib.indiana.edu/confluence/display/V3/

Definition of a FRBR-based Metadata Model for the Indiana University Variations3 Project. http://www.dlib.indiana.edu/projects/variations3/docs/v3FRBRreport.pdf

Definition of a FRBR-based Metadata Model for the Indiana University Variations3 Project, Phase 2: FRBR Group 2 & 3 Entities and FRAD. http://www.dlib.indiana.edu/projects/variations3/docs/v3FRBRreportPhase2.pdf

Variations3 Advisory Board Meeting 4/24/06 Metadata Breakout Session: Cooperative Cataloging Models

Variations3 Metadata Creation Guide. http://www.dlib.indiana.edu/projects/variations3/metadata/guide

Presentations

Variations Webinar presentation slides. http://www.dlib.indiana.edu/projects/variations3/oss-webinar.html (Recording and Q&A transcript also available online)

Publications

Dunn, Jon W., Donald Byrd, Mark Notess, Jenn Riley, and Ryan Scherle. “Variations2: retrieving and using music in an academic setting.” Communications of the ACM 49, 8, 2006, pp. 53-58. http://doi.acm.org/10.1145/1145287.1145314

Riley, Jenn, Casey Mullin, and Caitlin Hunter. "Automatically Batch Loading Metadata from MARC into a Work-Based Metadata Model for Music." Cataloging & Classification Quarterly 47, no. 6 (2009): 519-543.

Riley, Jenn. "Application of the Functional Requirements for Bibliographic Records (FRBR) to Music." In ISMIR 2008: Proceedings of the Ninth International Conference on Music Information Retrieval , edited by Juan Pablo Bello, Elaine Chew, and Douglas Turnbull, 439-444. Drexel University, Philadelphia, USA, September 14-18, 2008.

Riley, Jenn. "Moving from a Locally-Developed Data Model to a Standard Conceptual Model." Culture and Identity in Knowledge Organization: Proceedings of the 10th International ISKO Conference, 5-8 August 5-8, Montréal, Canada , edited by Clément Arsenault and Joseph T. Tennis, 124-130. Advances in Knowledge Organization 11. Würzburg: Ergon Verlag, 2008.

 


[5] “Editor’s Choice: iTunes Meets Wikipedia.” Science 313, 2006, p. 737.

[10] “On the Record: Report of The Library of Congress Working Group on the Future of Bibliographic Control.” http://www.loc.gov/bibliographic-future/news/lcwg-ontherecord-jan08-final.pdf