Themes

1. Strategies for our group:
   • Our Stakeholders may extend beyond NSF and the EC: seek natural allies
   • New focused programs and new parts of broader programs are both possible goals
   • Issuing a report by December is desirable: NSF will shape DLI 3 in mid-2003
   • 6th Framework networks of excellence may support transatlantic collaboration
   • Find ways to engage other NSF/EU groups (e.g., teleconference to their meetings)
2. Challenges:
   • Each audio channel is rigidly linear
   • Once you find something, how do you retain the access point you have found
   • Many uses of the spoken word raise complex legal and moral considerations
   • There is some case law on moral rights (in addition to copyright) in Europe
   • Preservation/sustainability standards and practices are still evolving
   • Synchronization of multitrack recordings and multiple representations
   • Mutable metadata may require special handling
   • Reliable cost estimates are difficult because so many things are emergent
   • Digitization is costly but unavoidable for existing collections
   • Authentication of segments and entire recordings requires technology and process
   • Dark archives remove retention disincentives, but prevent some desirable use
3. Approaches:
   • Virtuous cycle: better access facilitates more use which attracts more funding
   • A shift of focus from media management to file system management may help
   • Some form of resource sharing may be needed by small organizations
   • Short-term and long-term access may require parallel solutions
   • Risk analysis is a viable option when rights holders cannot be reliably found
   • Emergent networks (e.g., Freenet) may limit the scope of government action
   • OAIS, MPEG-7/21, METS may provide useful frameworks that we can build on
   • Leverage our spoken word expertise to improve some types of multimedia access
   • Research issues should include technology, infrastructure, and use
   • Facilitate access with non-speech audio (Malcolm Slaney, IBM Almaden)
   • Facilitate access with non-content features (e.g., gender, speaker change)
   • Leverage existing NIST annotation standards (e.g., annotation graphs)
   • Treat transcripts as data to be aligned using metadata (rather than AS metadata)
   • We should explore potential commonalities with data grids
   • Explore options such as summer schools to expand the pool of researchers
   • Consider both search, navigation and use together when thinking about access
   • Consider official languages, secondary languages, and spoken-only languages
   • Focus on preservability, but not on preservation
   • Recommend some holistic projects to explore integration issues
   • Provide centralized facilities for access and rights management (reading rooms)
   • Exploit distributed processing (SETI at home) and distributed annotation
   • Don't rely on demand digitization if you need to ultimately process everything
   • Estimate the quantity of spoken word materials to help explain the challenge
   • Adopt a strategy that balances research imperative with potential for social impact
   • Draw insights from the use cases to help shape the research agenda
   • Assure integrity with watermarks and auditable workflow