NSDL Logo
 Annual Meeting
 Swiki Main
Sunday Sessions
  New Projects Luncheon
  Orientation
  Posters
Monday Sessions
  Opening Keynote
  Intellectual & Economic
  Research Challenges
  DLs & Education
  Implications
Tu & Wed Strands
  Birds of a Feather
  Building Collections
  Deployment & Continuity
  Services Development
  User-Centered Design
  Committees
 

NSDL Services Interoperability and Web Services



Bill Mischo, University of Illinois at Urbana-Champaign


Dean Krafft, Cornell University

Kurt Maly, Old Dominion University

Michael Bieber, New Jersey Institute of Technology

Ed Fox, Virginia Polytechnic Institute and State University with Martin Halbert, Emory University

The session will provide a forum to discuss the types of services and protocols that are being developed or used in the NSDL, relate these to the growing web-services community, and discuss the needs of services characterization and discovery. The session will include some short presentations of some current service implementations to stimulate the discussion.


Notes - NSDL Services Interoperability and Web Services


Attendees: 32

NSDL services track has 37-39 grants with a lot of diversity of projects. A lot of the grants address interoperability issues.

Dean Krafft


What CI is doing with respect to services. Initial infrastructure has been central accessed through the nsdl.org. Search using SDLIP. Archiving is done by SDSC, but not as a really public available service. And initially, went with Uportal that people found wasnt what they wanted.

New mode of access to multiple portals, support of browser extensions, and better integration. For central MR, SOAP/WSDL, RSS access. Search over SOAP/WSDL, REST. NSDL.org will be a PHP based system for flexibility. Philosophy is open, lightweight mechanisms. Accessing the MR can be done via REST interface available now (see slides). Now considering what other queries should we support? Search engine style, but MR is structured, not full text. SQL? Xquery? Strawman proposal for some simple SOAP/WSDL, possibly REST. For the future, thinking about MR access for relationships. Committed to adding annotations, relationships.

MR Ingest need a lightweight alternative to OAI-PMH so have been looking at RSS and thinking about an RSS-OAI gateway. Search Access currently built over Lucene. Currently full text search design collapses multiple fields. SOAP interface will be forthcoming. Archive Access primarily currently through nsdl.org from the SDSC SRB. SOAP interface almost complete and should be releasable fairly soon. We will be putting up a description page of SOAP interfaces being developed. Want to play well with the web. Want to expose MR as a crawlable, indexable tree to enable google search. Expose relationships, support lightweight contribution and enable new user services.


Kurt Maly


Building high level services over various richness of metadata. Emphasis on automation (particularly with respect to OAI). When talking about federating through harvesting and you get the basic metadata, you have the problem of keeping services consistent while ingesting new data. Their federation has a number of services, some for humans, some for machines. Primary ones are search and discovery and are easy to handle with respect to consistency. Another service is searching by equation. Also have an extensive citation service. Also provide a search service as a web service.
Harvesting by OAI. Then implemented various post-processing mechanisms, e.g. to support the citation, equation services and subject resolver. These post-processing is automated, and important to the consistency of the collection. Citation processing can include extracting from the referenced documents to build up the relationships between records and objects. There is also an annotation service where annotations can be created linked to an object. Interface examples of citation service in slides. Equation processing includes identifying equations, filtering equations and storage. Subject resolver tries to extract possible subjects a paper is related to.

OAI as a harvesting protocol is a limited protocol. What if you want to do your own search? They have added a search API and a bookshelf service. Search service provides access to all fields as a web service over SOAP. The bookshelf provides a way for a user to have a personal collection. The two services talk to each other to allow for use in UI such as do a search, then add one or more results to the bookshelf.

We dont have to worry about performance yet at 450K records. The web services look promising for not just their federation, but for others beyond. Full automation still has some things to be hardened, and working to bring in Shibboleth.

Michael Beiber


Has a project for looking at Digital library service integration. Demo available (u:guest, p:guest). Search results can provide integration of further services. Example was to NASA site, so as you see a resource, further links (searches) are added to the view to provide a browse like way of moving through the information space. The initial links are done through structural analysis, not lexical analysis for navigation around the site. As you drill down, lexical linking then becomes available such as going to a thesaurus.

Integration through linking through structural and lexical. Looking toward bringing relationships too. After analysis of pages/sites, automatically generate links for applying in services. These can also be used to direct access content bypassing menu structures. They are also working on integrating a collaborative filtering engine over the coming months.

How to integrate DLSI into your system? Need to develop a wrapper (they help do this) to help identify the structural elements in the collection. Then develop some linking rules to related information and services (object to service rules). The integration engine will then go through and build the underlying link structure for the DLSI overlay on your collection. Third thing to integrate is to specify what thesauri or glossaries to use for the lexical analysis.

Issues: Reuse of service designed for one service that yould like to apply to another (peer review mechanisms are an example). Another issue is applying static services to dyanamic information.


Ed Fox and Martin Halbert


The digital library in a box is a suite of components that can be used individually, or in combination, to allow development of a digital library. Includes such components as search, recommender systems etc. Key concepts (developed by Hussien Suleman) are a componentized system (not monolithic). Use of lightweight protocols such as XAOI as an extension of OAI, which is a lightweight protocol.

OCKHAM is a project looking at the idea of developing a framework for a digital library. Wanted to look at developing some bridging type components that can be used in current DLs. Have four goals: develop a reference model (along the lines of OAIS), use that model to guide service development, evaluation and dissemination. OKHAM library network considers putting together a P2P network of web services (SOAP, WSDL, UDDI and others). Want to design this P2P network to be easily deployed into local libraries, or to be included in current library implementations. Testbed services to be developed include OAI-PMH to Z39.50 searching service, Alerting, browsing, conversation, cataloging and pathfinding services to start with.


Discussion


Libraries have been deploying services similar to DLSI through OpenURL? Beiber: Yes, been looking at OpenURL and they could integrate, via the wrapper, to allow external access to OpenURL. Discussion extended on this aspect of use of, and integration, with OpenURL.

Discussion of APS (physics) markup (related to Kurt Malys presentation). At the moment, theyre using gif images for some display, but this may relate to work in the MathML. But one of the problems is searching by equation through the source to develop the metadata records. The physics and math communities see the issues of equations, display, use etc. as an important area, but where there are problems. There was also some discussion on linking to actual papers and correct resolving of citations. There seem to be a number of projects dealing with these issues, so perhaps there needs to be a working group to address the issues.

Working with location based services. People often want to find information related to a location. Part of the problem is bridging the gap between GIS and some of the ocean/atmospheric data.

Issues of what google will crawl, for example only going down a number of layers in db9 implementations.

Tools needed for small collections where they may not want to develop, or use an OAI tool.

METS as a protocol (used in Fedora) as a simple metadata linking mechanism, which could perhaps be used for the GIS/gazetteer type info by linking in the other. Cornell bird project utilizes these relationships. METS converters are also part of OKHAM. California Digital Library is using METS for a content management system. The user is provided an interface that comes from a transform from a METS record as not only bibliographic metadata, but also structural metadata (e.g. leading to table of contents, paragraph locations etc.)

Is there a good differentiation of whats a service versus whats content (collection)? Characteristics of collections that are service-like, characteristics of services that are collections-like.

What about a registry of services for the NSDL? E.g. UDDI.

What about a registry of controlled vocabularies? Theres a lot of vocabularies being developed in various areas of the NSDL. Having some registry of vocabularies could allow for richer, more interesting services. Would need to have schemas developed and versioned for such vocabularies. Perhaps the lexical mapper (DLSI). Should there be a working group to address this area?



Comments

Please enter any comments in the following format.
  • (commenters' initials) - month/day [comment date]
  • comment





NSDL thanks DLESE for hosting the swikis for the NSDL Annual Meeting 2003.

 Swiki Features

ViewEditPrintLockReferencesUploadsHistoryMain PageChangesSearchHelp