The Science Education Resource Center at Carleton College (SERC) builds educational portals on an infrastructure suffused with digital library technologies and sensibilities. Yet our userslargely geoscience faculty looking for resources and information that will help them improve their teachingdon't think of us as a digital library (if indeed they have any notion of what that term means). Our portals identify focused topics of importance to educators and present information on these topics through a variety of structuresthere is no presupposition that a collection of metadata and a search engine will be the best solution. Yet as we built the tools to support the creation of these portals, it became clear that collection building (at least on a small scale) and, more generally, a systematic approach to collecting and organizing the information within our sites would be a recurring theme. These are traditional strengths of a digital library. So we built portal tools that integrate and interoperate with digital library elements at a number of levels.
This approach, building bridges between the educational digital library world and the broader set of technologies through which educational information is funneled, is not unique to our efforts. Many educational digital libraries are reaching across this divide in an effort to connect more directly with those they are trying to serve. The National Science Digital Library is incorporating Pathways projects that may build educational portals on the foundations of the extant central metadata repository and search services . Merlot has announced a partnership with course management system provider Blackboardtying their collection into the tools faculty use to connect with their students . A recent Digital Library Federation report "Digital Library Content and Course Management Systems: Issues of Interoperation"  highlighted the importance of this kind of effort. The potential benefits of connecting digital libraries more directly to the educational environment are clear. The details of the mechanisms through which this might be best achieved are not. The aforementioned report concludes, "we are at the point where some convincing demonstration projects are badly needed."
The portals we are building and the tools upon which they are built offer one such demonstration. As we built the technical infrastructure that supports our sites, we identified areas where digital library technologies or approaches could be profitably integrated. Our tools reflect a judicious choice of those digital library elements we felt offered the biggest payoff in making our portals easier to manage and more effective for our users. This article presents an overview of our portal technologies with a particular focus on our integration of digital library elements. It highlights one possible path connecting 'traditional' digital libraries to educational portals and begins to illuminate the benefits and challenges of such an endeavor.
About SERC Portals
SERC education portals follow a number of different models for pooling and disseminating community expertise on focused topics. For example, the "On the Cutting Edge" project  is built around a series of face-to-face professional development workshops for geoscience educators. Each workshop pools the expertise and energy of the participants to catalyze the development of collections supporting themes and topics explored in the workshop. At the conclusion of each workshop, its portal evolves from a workshop support website into a website that highlights these collections. For example, a workshop on teaching petrology was the launching point for a collection of exemplary teaching activities, a database of analytic instruments available in the community, and an extensive bibliography of petrology teaching resources from across the Internet.
The "Starting Point: Teaching Entry-Level Geoscience" project  takes a different approach: engaging editors on pedagogic topics relevant to educators teaching entry-level geoscience college courses (e.g., teaching with models, interactive lectures, field labs). Each editor pulls together community expertise to build guided introductions to the topic. These overviews serve as wrappers for collections of classroom-ready activities that exemplify each topic.
The websites that support these projects (as well as others at SERC) present their content in a variety of ways. A portal may include a mix of a simple linear set of pages providing a directed introduction to a topic, more loosely structured sets of pages each providing a jump off point to external resources, and search interfaces that enable exploration of larger collections. To support this diversity of information structures, we've created an authoring environment with a great deal of flexibility. The tool makes it easy for our authors to choose the most effective way to communicate their messageincluding drawing on digital library tools when appropriate.
The SERC CMS
SERC websites are built using a tool we call the SERC Content Management System, or simply the CMS . Our CMS is an integrated set of tools that allow site authors to create and edit our websites from within a straightforward web browser interface. Authors visit our administrative website, authenticate and then navigate to the editing interface for a given web page. There they can edit the page text within a standard web form text box. They format their text with a limited set of html tags supplemented with some custom tags, such as the tag to refer directly to a cataloged resource. The results are stored in a central relational database along with other editable properties of the page (navigation menu, page template, title, URL, etc.). The final site viewed by the public is generated dynamically from these elements.
This approach provides benefits both in terms of the technical quality of the actual html, as well as in the operational details of coordinating work among science educators from across the country. It provides an environment where the expression of content expertise is explicitly separated from the technical issues of constructing a high-quality website. The CMS ensures the resulting pages follow current best practices with regard to web standards and accessibility . It also allows for the intertwined larger scale problems of design, usability and information architecture to be addressed uniformly across the site without requiring the authors at each site to master these disciplines on their own.
Because the authors work entirely through the web-based interface, all the stages of site production can be coordinated centrally. The CMS provides versioning of site content, separate environments for development work and the live site, and central management of files and images used throughout the site. It eliminates the "who has the latest version of the file" problem that grows as multiple authors combine their efforts on a single site using traditional 'static' html pages. The CMS has proven to be an effective tool in enabling the collaborative authoring of our portals. Our CMS currently holds over 2500 pages of original content from a dozen projects. Some 70 editors working at 40 institutions across the country pulled this content together. These editors, many with little or no previous web authoring experience, drew upon direct contributions from hundreds of their peers in the science education community in creating our web portals.
Beyond the benefits of coordination and consistency that are available in many web content management tools, our CMS integrates tools and practices from the digital library community. At the simplest level, the CMS allows for the expansion of page metadata beyond the obvious information on title, URL, and author to include information like keywords and controlled vocabulary categorization. This additional metadata is exposed in <META> tags within the html and is used to drive our internal search mechanisms. Similarly, the CMS will make it straightforward to expose the sort of metadata called for by frameworks like the Semantic Web  that hold promise for enabling more structured information use throughout the web.
The next level of digital library integration is evident in how we handle links to outside resources. Each link is backed with a complete set of metadata describing the external resource. This metadata provides concise descriptive information aiding our users in evaluating prospective resources. It allows us to provide more effective resource discovery, and the metadata also serves as an integration point for a number of services. Ideally this metadata could be drawn from existing digital library collections. Our sites would act much as a traditional library special exhibit: choice items, exemplifying a focused topic, are drawn from the general collection and placed in prominent view. The existing digital libraries would act as the general collectionthe 'stacks' from which we could cherry-pick. In practice there isn't appropriate metadata available from outside collections for many of the online resources we'd like to reference. And so, for the near term, we've set up our own cataloging process, creating metadata for these resources and collecting them in small behind-the-scenes collections.
External references, identified by site authors or submitted by the community via forms throughout our sites, are moved into a cataloging workflow system whose end product is a complete metadata record for the resource. The first step in this system is to check existing digital library catalogs to see if appropriate metadata already exists. Many of our projects relate to Earth system education, so we coordinate metadata-sharing with the Digital Library for Earth System Education (DLESE). We use the DLESE OAI software to harvest all 14 collections currently aggregated by DLESE. This nightly harvest uses the OAI-PMH , which we also use to share our locally cataloged collections back to DLESE for dissemination throughout the National Science Digital Library  (see Figure 1).
Currently this effort pays off about 25% of the time; the metadata has already been created and can be harvested from DLESE. Our local catalogers do a simple one-click import (making the record available in the CMS) and their work is done. The balance of the time we have to rely on our own effortsbuilding a digital library behind the scenes. The resource continues through our workflow system, which provides online tools to coordinate the metadata creation. The workflow tools track cataloging progress, aid in distributing the work from different projects across our pool of catalogers, and record our quality assurance process. Most of our resources are cataloged to the ADN metadata standard , using copies of the DLESE Cataloging System  that we run locally. We also have cataloging tools for creating qualified Dublin Core, in line with NSDL requirements , as well as for capturing basic bibliographic information for print resources. The collections we build in this way are strongly focused on the subject of the particular portal and are of exceptionally high quality. Since our goal is to build an effective portal (rather than a comprehensive collection), we focus on identifying only the best resources (which in many cases fall under close scrutiny by authors and reviewers) rather than canvassing freely for resources to bulk up a collection. As these collections flow back into DLESE and NSDL, they represent a particularly strong set of resources within these larger repositories.
In-Line Referencing and the User's View of Catalog Records
Once web resources have been cataloged, they can be incorporated into the portals in a number of ways. The simplest mechanism is to insert a link to a resource within the normal flow of the text of a web page. Rather than coding a traditional URL, the author inserts an explicit link to the catalog record, making use of its unique internal id: (e.g., [resource 123]). When the page is displayed by the CMS, a link to the resource appears along with a small 'more info' link that points to an enhanced catalog record (see Figure 2).
This 'more info' link provides a view of the metadata record reduced to those fields of most interest to users (title, description, URL, author, terms from subject vocabularies). This concise synopsis of a resource is valuable for users in evaluating whether the resource might meet their needs. These pages also serve, unexpectedly, as a common entry point into our portals, as their simple, information-rich format makes them high-ranking targets for web search engines such as Google .
In addition to the core metadata, we augmented these display pages with various types of related information. Since each reference to a resource within our CMS is tracked, we are able to provide 'this resource is also referenced here' links so users can explore related areas within our sites by following these citations. We've also integrated third party services that provide information related to a given resource. For instance, the DLESE community review system  collects users' reviews of the teaching materials in the DLESE collections. We harvest the review status for DLESE resources (provided in a prototype annotation metadata format via the OAI-PMH) and, where appropriate, provide direct links to the reviews from our resource display pages. In addition, we take advantage of the archiving service provided through NSDL, linking to archived versions of the resources we reference so that users can investigate a resource even when it is temporarily off-line . The integration of these third party services is completely automated: users benefit from the direct linking of this information within our site without effort by our authors or catalogers. As similar services, connecting additional information to websites through catalog records, grow up around the digital library community, we'll be able to integrate them into this display page.
Linking to outside sites through a reference to a catalog record in this way also enables a number of efficiencies as we manage these outside links. For instance, link checking can be centralized and automated. If the URL of a referenced resource changes, we need only change the single central catalog record and all references to this site are automatically updated throughout the site. As record sharing becomes more widespread, these sorts of efficiencies are magnified. Updates made to metadata (e.g., a change in URL) by any of the DLESE collection providers automatically propagate throughout our sites with no human intervention.
Another way we make use of our cataloged resources is by building annotated bibliographies. Our authors frequently find the traditional annotated bibliography formata citation list of core resources on a topic annotated with brief, critical descriptions of each resourcea strong approach to presenting key references. The CMS provides a simple tool for aggregating and annotating groups of resources based on their catalog records. The resulting lists can be easily inserted into existing web pages. Authors may annotate the list with descriptions specific to that list or let the system fall back on using the existing generic description field from the metadata. As with stand-alone links, these annotated lists link to the resource metadata and the value-added services that can be tied to the catalog record . The ability to identify existing resources in peer digital libraries, and with a few clicks bring the resource into an online annotated bibliography, complete with all the benefits of backing metadata, exemplifies the value of working within a website authoring environment that incorporates digital library tools.
Building Collections of Community Expertise
Although identifying existing websites and print resources is a common way to build a collection of materials around a topic, in many cases valuable community expertise is not available in an easily reachable form. A great teaching activity may exist as only a set of handouts in a filing cabinet combined with the time-tested wisdom of the faculty member who uses it. There is often little motivation for individual community members to commit time to building web pages (or writing articles) to share their expertise on topics that fall outside the normal channels of scholarly exchange.
In an effort to tap into this knowledge, we've developed a process and set of tools for gathering this type of expertise. Educators are solicited to fill out a relatively simple web form that has been carefully structured to tease out the information needed to make that bit of knowledge more easily reused. For instance, we collect educational activities related to petrology, asking not only for a description of the activity but also for the course context in which it's taught, learning goals, assessment, and general teaching tips. Educators can also use the integrated upload tool to provide relevant files (e.g., the Word file they print and hand out to students, the Excel spreadsheet that contains the base data, etc.) . The form is built around the elements we've identified as necessary to enable other educators to adapt and use activities successfully in their own environment.
The CMS provides a simple online tool for building the collection form (often just a matter of building a variation on an existing form) and embedding the form within an existing page. Submissions automatically collect in an online queue where they can be vetted (e.g., for duplicate or incomplete submissions and appropriate use of controlled vocabularies), and then with a single click the submission becomes a new web page within the portal. These pages aggregate into 'mini-collections' that, by virtue of the topical categorization the submitter has provided, can be automatically exposed in search/browse interfaces .
We can follow this same model to aggregate community knowledge beyond teaching activities. For example, we currently use these tools to build searchable mini-collections of geoscience course syllabi (with associated information about the design of the course) and a registry of geochemical analytic instruments. While the tools themselves don't eliminate all the challenges in bringing together this sort of community expertise, they do reduce several critical hurdles. The tools minimize the effort needed on the part of the contributor to add their expertise to a community pool; contributors need only fill out a web form and upload any existing supporting files they have. The tools also drastically reduce the editorial effort needed to transform the contributions into a coherent collection; a single click turns a contribution into a new page in a searchable collection. By minimizing these logistical barriers, we hope to empower communities of educators to collect expertise and insights previously unshared.
When a site author sees the need to create a collection of resources (either cataloged references to external websites or content gathered through the site into a mini-collection), the CMS provides an easy mechanism to set up search interfaces specific to the collection. Our current interfaces are based on ideas explored in the Flamenco project's work on faceted search  ideas that are starting to show up in other digital libraries . The central theme is to simultaneously provide both a full text search interface and a browse interface over multiple controlled vocabularies. Users can freely combine text searching with browsing, employing either technique to further refine a search.
The search screen provides a search text box at the top with search results below (see Figure 3). On the right side of the screen, the controlled vocabulary terms are displayed. Choosing a particular term limits the result set to the part of the collection matching that term. Each vocabulary term is displayed along with the number of resources matching it. This number updates dynamically, reflecting the user's current search text and browse choices. This provides immediate feedback about the scope of each search as well as providing direct access to deeper browsing through hierarchical vocabularies. The current search parameters are listed at the top of the page, providing a simple mechanism for expanding as well as narrowing a search. The controlled vocabularies serve double-dutyas both a browse interface and a feedback tool for search scope.
The underlying technologies are those found in many digital libraries. Our full-text search uses the Lucene library , implements Porter stemming , and has a search syntax designed to match our users' most familiar touchstone: Google. The browse structures displayed are a natural expression of the controlled vocabularies used in our metadata. However, explicitly exposing the browse structure with its integrated search feedback alongside the full-text search provides users with a more direct sense of the information space they are moving through, empowering them to explore more freely. Research results indicate that this sort of faceted search interface can lead to more successful search experiences than some of the more traditional options .
The success of an interface like our search interface is highly dependent on the efficacy of the underlying controlled vocabularies. While we are making use of a number of existing vocabularies, we've found it important to develop specialized controlled vocabularies to fit individual collections, or sets of collections. These specialized vocabularies have the right granularity and particular choice of facets that do the best possible job of exposing a particular collection. To support broad and flexible application of controlled vocabularies, our CMS allows point and click creation of new vocabularies (flat or hierarchical) whose terms can be applied to any catalog record or individual web page within our system. This low barrier to creation and application of new vocabularies makes it easy to quickly explore the use of a new vocabulary in a small collection. This in turn enables rapid improvement of vocabularies based on evaluation of actual use.
An integrated indexing system builds a continuously updated full text index of all web pages and catalog records, exploiting their structured nature to extract vocabulary information and using variable weighting of different page elements to improve search results. The CMS also supports various manipulations of web page specific metadata to enhance local search effectiveness, including page keywords and arbitrary boosting of pages that site authors indicate are of especially high quality on a given topic. The CMS provides a set of straightforward yet powerful tools with which our authors build and fine-tune vocabulary-driven interfaces for exploring their collections.
SERC educational portals leverage the techniques and tools of the digital library community to bring cohesion and clarity to their presentation of information. We make wide use of the rich descriptive metadata and services tied to our collections, taking advantage, where possible, of overlapping efforts elsewhere in the community to share metadata and services. We build collections reflecting community wisdom in ways that are systematic, scalable and driven by the real needs of their final users. We exploit this order and intent through interfaces that implement current best practices enhancing users' exploration of our sites and collections. We hope that our CMS serves as a demonstration of what is possible when the tools for building informational websites are built on foundations that express the experience of the digital library world, organizing information in ways that help users better explore, evaluate and share the knowledge of their discipline.
For those looking toward similar sorts of integration between digital libraries and educational delivery tools like Course Management Systems, there are several key observations. First, it can be done. Our authors are happily building bibliographic lists within a web-authoring tool that automatically draws the metadata from digital libraries. Second, while were able to exploit existing library content from closely aligned digital libraries, there is still (at least in the areas that our users are working) a significant body of material not available in existing repositories. Having the ability to create new metadata to describe resources from the larger web, as well as to develop the tools needed to bring original materials online, is crucial to our site's achieving its goals. The path is not a one-way path from digital library into educational tool; it is a loop, perhaps tighter and with a different asymmetry than one might expect. Finally, the tools and standards enabling interoperability are still in their infancy. Ideally, a portal builder might be able to rely on search and display tools from existing digital libraries whose interfaces integrate with the portal in real time. But our digital library integration is largely dependent on asynchronous data shuffling via OAI-PMH, with search and display handled by locally built and installed tools. This has been, for us, the most expedient way to reach our goals of building effective educational portals. We'll know progress has been made when we can assemble portals of similar quality and functionality with a much simpler layer of digital library 'machinery' embedded in the portal.
This work is funded by the National Science Foundation under the grants: DUE 0127141, DUE 0226243, DUE 0226199, and EAR 0304762.
Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.
Notes and References
 National Science Digital Library program solicitation, <http://www.nsf.gov/pubs/2004/nsf04542/nsf04542.htm>.
 Blackboard press release describing Merlot partnership, <http://www.blackboard.com/about/press/prview.htm?id=608612>.
 CMS is a terribly overburdened acronym within the educational technology communityeven getting beyond the ambiguity over whether the 'C' stands for 'Content' or 'Course'. We beg the reader's indulgence in our use of the acronym as it has become standard parlance within our office, and retrofitting a new name at this point seemed artificial.
 The term "web standards" in this case refers largely to adherence to the World Wide Web Consortium (W3C) standards, (http://www.w3.org/), as promoted by the Web Standards Project (http://www.webstandards.org/). Adherence to these standards is intimately tied to creating sites that are as broadly accessible as possible. See, for example, AccessNSDL (http://accessnsdl.org/) for accessibility information related specifically to digital libraries.
 Open Archives Initiative Protocol for Metadata Harvesting, <http://www.openarchives.org/OAI/openarchivesprotocol.html>.
 NSDL metadata requirements are described in the NSDL metadata primer, <http://metamanagement.comm.nsdlib.org/contributing.html>, and specifically in the xml schema NSDL provides, <http://ns.nsdl.org/schemas/nsdl_dc/nsdl_dc_v1.02.xsd>.
 In November 2004, 13% of visitors arrived at our sites through a catalog record page.
 Moore, R., "Preservation Environments," NASA / IEEE MSST2004, Twelfth NASA Goddard / Twenty-First IEEE Conference on Mass Storage Systems and Technologies, April 2004.
 For one example, see <http://serc.carleton.edu/NAGTWorkshops/visualize04/recommend.html>.
 Hearst, Marti, Jennifer English, Rashmi Sinha, Kirsten Swearingen, and Ping Yee, "Finding the Flow in Web Site Search," Communications of the ACM, 45 (9), September 2002, pp.42-49.
Copyright © 2005 Sean Fox, Cathy Manduca, and Ellen Iverson