Paths over the World-Wide Web: Customizing Services

Unmil Karadkar, Luis Francisco-Revilla, Richard Furuta, Frank M. Shipman III,

Avital Arora, Emily Luke

Texas A&M University

Center for the Study of Digital Libraries and the Department of Computer Science

H. R. Bright Building, College Station, TX 77843-3112, USA

Phone: +1-979-845-3839, Fax: +1-979-847-8578

{unmil, l0f0954, furuta, shipman, avital, eluke}@csdl.tamu.edu

Contact Author:

Unmil P. Karadkar

Texas A&M University

Department of Computer Science

408-F, H. R. Bright Building

College Station TX 77843-3112

USA

Phone: +1-979-845-4924 e-mail: [email protected] Paths over the World-Wide Web: Customizing Services

Abstract

Walden's Paths enables the recontextualization of Web-based documents by implementing the organization structure of linear directed paths. Paths, long-recognized as a useful adjunct to hypertext, overlay a structure over pre-formatted information without modifying its original structure and presentation. Walden's Paths can be thought of as meta-documents that point to other documents on the Web. Through paths, one can contextualize, elucidate, annotate and communicate previously independent Web information to meet the needs of an intended audience. As the scope of the audience increases, however, their individual differences require customizations to paths in order to address specific needs, preferences, and abilities. This paper describes some of the customization mechanisms that we have added to Walden’s Paths in order to allow specification of paths that support wider audiences. Specific mechanisms discussed in this paper include support for multiple languages, for virtual collections of paths, and for paths that are created on-the-fly to reflect a user’s characteristics, perhaps representing a tailored version of a previously-specified path.

1 Introduction

Over the decades, visionaries and researchers have recognized and elucidated the usefulness of

(linear) paths as a means to organizing and sharing information. Vannevar Bush was the first to point out the value of creation of a path over pre-organized content (for example, books) to reorganize and associate related information for easy retrieval and communication (1).

2 Subsequently, paths have been incorporated into many hypertext systems, most significantly in

Notecards (2) as Tabletops and Guided Tours (3) and in Active and Scripted Paths (4), (5).

Walden's Paths (http://www.csdl.tamu.edu/walden/) is a World-Wide-Web-based implementation of the path mechanism, enabling the overlaying of a linear, directed structure over Web information resources. These paths are meta-documents that point to other documents on the Web—paths federate (6) physically separated information on the Web into a coherent collection. A path does not modify the documents that it points to, in either structure, content or presentation, nor does it restrict access to the original documents. Authors of paths are individuals with domain knowledge who add contextual and encapsulating information to the

Web material. In general, the author of a path is not the author of the information that the path threads together and consequently has little or no control over either the content or presentation of the information.

Walden's Paths was originally implemented for use by K-12 teachers and students. In this use, teachers use the system to contextualize and convey information by creating paths that students browse. We have described our experiences in this environment in more detail earlier see (7),

(8), (9), (10)—an additional use of paths that we observed was as a means of communicating information from students to teacher (e.g., to present information discovered during a term project) and from student to student (e.g., information prepared by a more-advanced student for use by others).

3 As we have continued to explore the uses of Walden’s Paths, additional applications have been suggested involving broader groups of users. For example, we currently are investigating the user of Walden’s Paths as a means of providing overviews of an already developed general- interest collection for different classes of viewers (e.g., school children, university undergraduates, graduate students, researchers, and members of the general public) and in multiple languages.

Section 1.1 briefly introduces Walden’s Paths and section 1.2 reviews the use context to be addressed by the features described in this paper. The subsequent sections describe the features that have been provided to support customization. Section 2 describes extensions to the interface and authoring tools in support of multilingual viewer populations. Section 3 describes virtual directories, a means of providing different views of a collection of paths to different groups of viewers. Section 4 discusses ephemeral paths, which is a mechanism that allows the path itself to be adapted to different viewers’ characteristics. Section 5 contains discussion and directions for future work.

1.1 Walden's Paths

Authors of the paths include the information they intend to convey to the viewers by pointing to resources (articles, illustrations, media files, etc.) on the Web and adding their own comments to express this information in context of the subject matter at hand. (In this discussion, we will classify the users of Walden's Paths as ‘authors’ who create the paths to share information and

‘viewers’ who use these paths.) A stop on the path is called a ‘page’. The author may also add his contact information, the dates of creation and expiration and a brief overview of the topic that

4 the path addresses. Authors perform duties that are similar to those of publishers and editors of books. A Publisher reads many manuscripts and selects the ones that, in his opinion, will be valuable to the viewers. An editor modifies the manuscript to make it suitable for public presentation. Finally, the publisher confirms his faith in the manuscript by investing the money to publish and market the book. Creators of the paths act as editors and publishers of Web documents. They carefully sift through information elements and bestow legitimacy on the ones that contain the most relevant information by including them in the paths and by providing their contact information as stamp of their approval.

In K-12 classroom sessions as well as workshops conducted for evaluation of Walden's Paths, we have observed that the metaphor of the path is intuitive to students as well as teachers (11). After using the system for a short time, teachers and students quickly focus on contents of the path.

The technology vanishes into the background allowing users to concentrate on their primary task of understanding the contents of pages.

Walden's Paths consists of three distinct programmatic units. Authors create paths using the

'Walden's PathAuthor', a Java based application and save them in the local storage space (for example, hard disk). The authors must then ‘publish’ these paths to a Web accessible location before the viewers can access them. ‘Walden’s PathPublisher’, a CGI application stores paths in the authors’ directories on the server and makes them available to the viewers. Viewers may view the published paths using any standard Web browser that supports frames by pointing it to another CGI application called the 'Walden's Path Server' (also referred to as 'Path Server' or

'Server'). This paper focuses on the customization features provided in the Path Server. Section 2

5 briefly overviews the authoring tools. For a comprehensive discussion of the authoring process and tools, see (12).

Figure 1: Walden's Paths start page

Figure 2: Information page for a path

6 Viewer interaction with the Path Server begins with a list of available paths. Figure 1 shows the typical view of the start page of the Path Server. Viewers can opt to see the abstract and a listing of pages in the path by clicking on the ‘i’ (for information) icon. Figure 2 illustrates the

‘information view’ of a path (that is not included in the list of paths in Figure 1).

Clicking on the title of any path opens it and displays the first page on that path, as shown in

Figure 3. Most of the viewer interaction with Walden's Paths takes place in this mode. Viewers see three frames. The bottom frame displays the resource that the path points to exactly as it would appear if viewed via a standard Web browser without the Walden's Paths interface and is called the ‘Content Frame’. The top-right frame, called the ‘Annotation Frame’ displays information added by the author for contextualization of the resource. The top-left frame, called the ‘Control Frame’ contains widgets for interaction with the Path Server. Following the link associated with the ‘Walden's Paths’ image takes the viewers back to the list of paths that they started from. The left and right arrows to either side of the image allow users to navigate backward or forward along the path in order of the pages. A listing of the neighboring pages facilitates out of order traversal of the path. Viewers can access the resource that the path points to, and view it outside the Walden's Paths interface by clicking on the linked URL of the resource.

Viewers can follow links contained in the Web page that the Content Frame displays. Doing so takes them ‘off the path’. Figure 4 displays the state of the interface after a viewer follows the

‘Facilities’ link in Figure 3. The Control Frame then displays a ‘Return to Path’ image that

7 allows the viewers to return to the path where they stepped off with a single mouse click, as shown in Figure 4 and resume traversal along the path.

Figure 3: A page on the CSDL path

Figure 4: Going off the path

8 1.2 User level Customization

The domain experts create paths that address the overall needs of the target population. However, not all users within this population are identical. They may vary over dimensions other than the ones that define the population. Interaction of a particular user may also vary due to change in intent as time passes. For example, K-12 students share some properties with each other, in that they are a young audience, attend school and study subject matter. Yet, they vary drastically over other characteristics. There is a wide age difference across the spectrum, they learn various subjects from different teachers and may access the paths for a variety of reasons, for example, to understand a topic as it is covered in the class or to revise it before a test. The Path Server facilitates viewer customization of information to suit their individual characteristics and changing needs. The authors can also use the customization features to address the needs of various user groups. Students in elementary schools, for example, may need assistance from the teacher in customizing their environment.

This discussion assumes the most general setup in the Web scenario, where authors across administrative domains and subject areas share a Path Server to address users who prefer diverse languages. In the K-12 scenario, all schools in a school district that provides bilingual education may share a Path Server to conserve resources. Similarly in the context of digital libraries, reviewers may link their comments to the digital document (or a particular page in the document) that they address. Researchers could use paths to tie in bodies of related research work to form a coherent presentation of their background study. With many authors creating paths to address their audiences, the generic list of paths as shown in Figure 1 will soon grow to a size where it will be tedious for the viewers to find relevant paths.

9 Viewers can customize their environment at various levels using the Path Server to view only the relevant information. They can specify criteria that display lists of paths that are relevant in their current context and tailor paths to view only the pages that are of immediate interest to them.

They can also customize the Walden's Paths environment to access information in a language of their choice. We currently support three languages: English, Spanish and Hindi.

Many customization and personalization models require users to fill in profile information. My

Yahoo! (13) allows users to customize content and presentation based on user profiles.

Amazon.com (14) provides a personalized recommendation system based on the users' earlier purchases to better suit user preferences. Keeping in view the young age of our primary audience, we have refrained from creating profile-based user-specific models for personalization of the system. Though this restricts personalization that suits each user, we have been able to address many of our concerns.

2 The Walden's Paths Multilingual Interface

With the proliferation of the Web, one can expect that viewers of Paths come from a variety of countries and speak different languages. Moreover, some collections inherently attract multilingual audiences—for example the Cervantes Project’s Web site

(http://www.csdl.tamu.edu/cervantes/) maintained in our Center must provide for English-only and Spanish-only speakers as it chronicles the works and history of the author of the well-known

Hispanic masterpiece Don Quixote.

10 The Path Server offers the viewers a choice of viewing paths in various languages as a first step towards addressing non-English speaking audiences. We currently support three languages:

English (the default language of the Path Server), Hindi (the national language of India), and

Spanish (15). The degree to which a path is accessible in a particular language depends on the degree to which the path’s author provides support. It is straightforward to alter the characteristics of the ‘skin’—to provide navigational support in the appropriate language.

However, fuller support of the viewer can be provided when the author provides translations for page annotations. When the source material also is available in multiple languages, it is appropriate to allow the specification of the corresponding language version. However, in all these cases, the path infrastructure also must deal with the case when the author does not specify alternatives, possibly because appropriate alternatives do not exist or perhaps because the author is not fluent in a particular language.

Authors, then, need to create paths in multiple languages. This involves effort at two levels. The authors must find relevant Web content in the languages that they wish to use. This will be gradually easier as users of the Web add content in diverse languages. The authors must then annotate these resources in the corresponding languages. One can think of applications where annotating pages in a language other than that of the Web page it points to might serve a purpose. If an author is not conversant with all languages that he wants to include in his paths, he can collaborate with other authors who are fluent in these languages.

11 Figure 5: The Multilingual Path Author

Figure 5 displays a snapshot of the PathAuthor taken during the process of authoring a path. The

PathAuthor displays information about the Path on the left, that is, the title of the Path and an ordered listing of pages. The authors can title the path differently in various languages. Currently the PathAuthor only supports page titles in English. The ‘Up’ and ‘Down’ buttons reorder the pages by shifting the highlighted page above or below in the list. The right part of the display contains the information for the current page. The author may provide a title, the URL for the

Web document that the page points to and an annotation for each of the languages supported by the path. In view of the difficulty in finding corresponding Web pages in different languages, the

PathAuthor does not enforce completion of all fields for all languages. The right part of the interface also contains widgets for insertion, deletion and navigation of pages. The ‘Undo

Changes in Page’ button allows authors to revert all changes made to the current page since its display. The PathAuthor supports the standard ‘New’, ‘Open’ and ‘Save’ file operations. The

12 Figure 6: The PathPublisher authors may also provide their contact information, creation and expiration dates and an abstract for the path. This information is displayed when the viewers view the path information by clicking on the ‘i’ icon (as illustrated in Figure 1).

Figure 5 displays the path title in Hindi and the page details in Spanish to highlight the widest range of features. The Annotation field accepts HTML source or plain text as its input (notice that ‘ñ’ character in Spanish is written as its HTML equivalent, ‘ñ’). A path that contains the page information for all languages that it supports is called a ‘master path’. This path serves as a model for creating individual paths in each of the languages. The relationship between the master path and a path generated from it for a particular language is similar to that between a

Class and an Object (an instantiation of a Class) in Object Oriented Programming. Paths for

13 different languages are generated from the master path that serves as a conceptual and semantic model for the paths.

Figure 7: Page on a path in Hindi

Figure 8: Page on a path in Spanish

14 The PathPublisher separates a ‘master path’ into its ‘constituent paths’ automatically while publishing it, one for each language supported. Doing so reduces the processing required by the

Path Server, but also makes switching between versions of a path more awkward. Figure 6 displays the interface of the PathPublisher. The PathPublisher employs user authentication to prevent unauthorized tampering with the authors' published paths and saves them in the authors’ directories. Authors cannot modify published versions of a path. Instead, they must edit the path locally, delete the published version of the path, and then republish the modified path. This method, while cumbersome, ensures synchronization of different paths generated from a master path. Parallel pages in a master path may not be available in all languages. This missing page must be filled in for maintaining the sequence of pages and flow of narrative across the paths.

The PathPublisher addresses this issue via the use of the ‘default’ language. The corresponding pages in the default language replace all missing pages in any of the languages.

Figure 7 displays the 4th page in the Hindi version of the ‘Travel India’ path. This page describes the ancient Buddhist Stupas at Sanchi, located in central India. Figure 8 displays the corresponding page in the Spanish version of the path (notice that the path title in the Control

Frame is in Spanish). However, as a page on Sanchi was not available in Spanish, the

PathPublisher replaced it with its counterpart in the default language. Hence, the annotation, as well as the content appears in English. Such pages appear discordant and the authors must address the issue to spare the viewers of the path from unexpected changes that may distract them from focusing on the page content.

15 The discussion about the authoring of multilingual path discusses only the creation of parallel paths. A word of caution is in order here. The parallel paths represent the most general structure of a path and account for the possibility that not all information may be available in a particular language. Authors may choose to create a path that contains all pages in a single language. This is a special case of a parallel path. The creation of parallel paths also presents a possibility of using these in language instruction. Viewers of paths may switch between languages in a parallel path, one that they are familiar with and the other that they wish to learn.

As a first step towards internationalization of Walden's Paths, viewers can view paths in one of the supported languages. Multilingual paths extend the benefits of contextualized, authoritative information to non-English speaking viewers. Though authors of these paths can annotate the pages in various languages and scripts, the interface of the PathAuthor and the PathPublisher does not support other languages. (16), (17), (18) document the technical, social and cultural issues in designing software for international audiences.

3 Virtual Directories in Walden's Paths

In real-life situations, many authors will need to share collections of paths—perhaps because they are working on partially overlapping projects or perhaps because they work for a common administrative entity that provides only a single installation of the Path Server. Each of these authors addresses a subset of the audience of the collection—some paths may be used by multiple groups, others by a single author’s group. In the distributed version of the Path Server, designed for the uses of a single author or a small group of authors, a single directory listing is provided for each Path Server installation. However, as the number of paths published in the

16 collection increases, viewers find it increasingly difficult to find relevant paths due to both the quantity of the paths as well as variety in the topics that they address. Our first attempt to alleviate the situation explored the possibility of exposing the internal path organization by displaying path lists for authors. Thus, the main path list would display links to the path lists for various authors. In Figure 1, the last few entries point to lists of paths created by authors.

Viewers may access an author's path list by selecting the corresponding link, just as they would access an individual path. This resulting list of paths contains only those created by the particular author and is visually similar to the list shown in Figure 1. We soon realized that this design was inadequate as authors created paths for different groups of users (within the target population) over different topics and the interests of viewers changed periodically.

To see this, consider a K-12 scenario. Here, teachers teach multiple subjects to different grades and in some cases, at more than one school. Figure 9 (19) shows an example that is possible in a real life K-12 school district. In the situation diagrammed, Mrs. Jones teaches Social Studies to

8th grade students at two schools, A and B. Mr. Smith teaches Math and Science at school B to the 7th and 8th grades. Figure 9 shows possible directory locations for paths created by these

Figure 9: Various domains for teachers and students

17 teachers for each of the classes they teach. It also shows the schedule for a student in the 7th grade at school A. The domains and hierarchies that apply to the various students and teachers may overlap and even span administrative boundaries (in the case of Mrs. Jones). Thus, no fixed hierarchy of static directories can yield a listing of all paths that are relevant to every viewer while satisfying the authors’ expectations of privacy and security. Content-based access offers a solution in such scenarios.

3.1 Virtual Directories

In most file systems, users access the data by exploring a fixed, hierarchical directory structure.

They must remember locations of the files in terms of a possibly intricate, idiosyncratic and non- intuitive nesting of directories. Over the past decade, file systems like Semantic File System

(SFS) (20) and the Essence system (21) have successfully delivered content-based access. These allow users access to large amounts of data while freeing them from the need to remember elaborate file paths and directory hierarchies. The Hierarchy and Content (HAC) file system combines the concepts by providing simultaneous content-based as well as hierarchical access to files (22).

3.2 Virtual Directories and paths

Virtual directories address the challenge of displaying relevant paths to all users independently of their expectations. Thus, viewers as well as authors can create virtual directories and view all relevant paths as a single list, irrespective of the location of the paths in the hierarchical author- based structure (implementation limitations in the current version requires that the paths be served by the same Path Server). The paths representation now includes data stored as attribute-

18 value pairs and the elements of a virtual directory are selected by matching desired values.

Viewers can specify the terms to be matched and the attributes to be searched—in many applications we also would provide a pre-selected set of values (e.g., to make selections relevant to ‘Spanish speakers’ and ‘English speakers’; or for ‘college students’ and ‘lifelong learners’).

To support the K-12 example presented in this paper, the set would include attributes like grade level, authors, subject, date of creation, keywords and the text contained in the annotations. The keywords include terms that are related to the page or the path but do not appear within the annotations (19).

Virtual directories provide another crucial service for rapidly changing systems. The system stores parameters that the users provide and resolves them at access time. The resultant set of documents reflects the most recent system state and returns only the paths that are currently available for viewing. Authors can aid in setting up virtual directories for user groups with special needs by inclusion of specific values for the various parameters.

3.3 Creating and using Virtual Directories in Walden's Paths

The Path Server allows users to create virtual directories by entering query terms to match against paths in the system. The system saves the query terms provided by the user are returns a

URL that acts as a handle to the virtual directory for future accesses. When a viewer accesses the handle, the system resolves the query and returns the results as a single list. Users may modify virtual directories by changing the query terms and criteria associated with them. Users can choose to overwrite the existing virtual directory with the new criteria or save them as a new directory. The next subsection describes user interaction with the virtual directory interface.

19 Figure 10: Virtual directory creation interface

The virtual directory creation interface is shown in Figure 10. The interface supports ‘basic’ and

‘advanced’ query modes. In the basic mode, the users may type in keywords that they wish to match against the paths. The system returns the paths that match these keywords in any attributes. In the advanced mode, viewers specify the terms as well as the path attributes to match these terms against. The Path Server searches for the query terms only in the fields specified by the viewer. This mode gives viewers a finer control over selecting paths that they wish to view.

In either of the modes, the creator of the virtual directory may choose to include a link that points back to the main path list (shown in Figure 1).

20 Figure 11: View of a virtual directory

Figure 11 depicts a virtual directory that links back to the main path list and displays all paths that contain the terms ‘Music’ or ‘Art’. Accessing the paths from a virtual directory is identical to accessing paths from the main path list or any of the static author path lists. The location bar of the browser displays the URL that acts as the handle to this sub directory. The handle contains a 16-character identifier that uniquely identifies a virtual directory. The system enforces the uniqueness of the handle during its creation. Viewers may bookmark the handles (they are URLs, after all) and return to these at will.

Users can modify a virtual directory via a link in the top right corner of the browser window. The interface for modifying virtual directories, shown in Figure 12, is similar to that for creation of a virtual directory. The system displays the current parameters of the virtual directory to assist the viewer with modification of the these parameters. Viewers may choose to overwrite the current virtual directory with the modified one, or save the modified parameters as a new directory. The

21 Figure 12: Virtual directory modification interface interface modifications affect only for the current mode (basic or advanced) of the virtual directory. The interface for the other mode remains unchanged.

Users may search for virtual directories using the interface for creating virtual directories. In this case, the system does not create a new directory, but only searches the existing virtual directories with matching criteria and returns a list of these to the viewer. The basic search, as expected, returns all virtual directories that contain any of the keywords specified in the search within their criteria. An advanced search would return only directories where the search terms match within the specified fields.

The inclusion of additional attributes in the paths improves the quality of metadata associated with them. This directly results in a value gain as richer metadata structures can support diverse services over the collection of documents (in this case, paths). Currently all metadata associated with the paths is created manually. Automating some of this will ease some load on the authors.

However, it is advantageous to have a human element in the creation of metadata as it allows for providing context based on the nature of materials or their use apart from just the terms that make them (23).

22 4 Ephemeral Paths

In addition to customizing the list of available paths, it also is useful to enable the customization of the contents of individual paths, perhaps to allow viewers to concentrate on specific pages within a path. Viewers access paths with different intentions. They may initially view a path to study and understand its contents and may view all pages on the path. However, when they return to paths that they have accessed before, they may be interested in specific pages. Users can create sub-paths from existing paths for short-term use (11). The process of creation of a temporary, ephemeral path is easier than the author-publish-view process described in section

1.1. To offset the low long-term utility of the paths created via this mechanism, the creation of these short-term paths must itself be a lightweight process. The Path Server includes an interface for easy creation of ephemeral paths via selection of specific pages on an existing path. This process of selecting a subset of the pages on a path to create a lightweight path is termed as ‘Path elision’. Users may create sub-paths of pages that interest them, those that they find difficult to understand or for a variety of other reasons.

23 Figure 13: Interface for path customization

Figure 13 shows the interface for creating an ephemeral path. This interface displays metadata about the path and a listing of all pages in the path (similar to Figure 2) along with a checkbox next to each page. A viewer can select pages to be included in the ephemeral path by checking the corresponding boxes. The user selects the pages to include in the sub-path and clicks the

‘Create Customized Path’ button. This results in the creation of a path that consists of the selected pages. This path is not included in a list of paths as it is not intended for long-term use.

Instead, the URL of this path acts as a handle to access it (similar to the virtual directories) and

24 can be book marked for later access. The system may remove transient paths periodically. Figure

14 shows the ephemeral path created from the path shown in Figure 13.

The uses of ephemeral paths are broader than simply the customization of preexisting paths. We also are investigating use of ephemeral paths in other situations that call for transient access to information (11), for example, to generate random tours or previews of digital collections from art galleries or museums. Paths that access transient materials is another case where this mechanism will be useful. In this case, the path structure may include documents, the contents of which change frequently, yet the underlying structure is persistent. Web sites of news services and weather forecasts are the most representative of this class of documents.

Figure 14: The ephemeral path created in Figure 13

25 5 Discussion and Conclusions

In its present form, the Web lacks crucial services of cataloging and collection development among others (24). It does not support the vital roles that editors, librarians and publishers of books perform to ensure the trustworthiness of materials viewed by the users (25). Walden's

Paths provides a service that helps people with domain knowledge reformulate Web-based material into a digital library context by supporting library services and human-oriented federation of Web materials. Authors of paths add metadata to Web-based information. This human-generated metadata aids resource discovery and is useful for recording within-collection organization (23). The success of the Yahoo! catalogue is a testimony to the value of human element involved in the cataloging process (26).

The quality of services that Walden's Paths provides can be enhanced by extending the system along three broad directions: expansion of its customization features, support for management of rapidly changing distributed digital collections (27), (28) and exploration of the value of richer path structures. In this paper, we have discussed mechanisms directed to the first of these directions.

An international version of Walden's Paths with support for authoring paths in various languages will expand the community of authors and provide an impetus to authoring of multilingual paths.

It will also provide a testbed for studying issues in supporting collaborative authoring of paths between creators who may have little common ground due to educational, social and cultural differences.

26 Just as information must be customized for use by diverse populations, the customization tools must themselves be targeted at the experts in various domains. The needs of the experts vary by their domain, technical orientation, experience and most importantly, the time they can spare for this task. Thus, the tools that are appropriate for K-12 teachers for creating paths, may not be suited for use by domain experts in providing library services. The diversity in needs of domain experts must form a basis for the tools that developed to aid them in creating and viewing the collections.

As we expand our audience to include adult and lifelong learners in addition to the users of digital libraries, use of simple profiles for personalization deserves serious consideration. In educational and research environments with defined, tangible goals, it may offer value than in generic domains with goals that either vary with time or are not very well understood. For example, when personalizing an environment on My Yahoo!, users need to choose from hundreds of modules and to fill in long profiles (29). My Yahoo! caters to users’ daily information needs and the domain knowledge is not expressible in tangible terms.

The Web is an enormous, dynamically changing, decentralized, unregulated information repository (30). We are working on providing solutions to manage collections over repositories that, like the Web may change in an unpredictable manner and have an initial prototype of a

PathManager (27). Effective management of collections over the Web requires an understanding of how creators of the paths perceive changes in the Web content. Some initial results are presented in (28). A further understanding of the nature of changes on the Web, and the perceptions of authors of the path will influence future development of the PathManager.

27 The simplicity of linear paths is an asset for their practical use in many environments, including educational environments. Non-linear tree or graph based path structures could be used for a multitude of purposes. For example, a tree-based path structure that follows the progress of a research project. Branches of the tree may map to the hypotheses that the research had tested and the results of each. However, the use of non-linear path structure presents visualization challenges for authors as well as users.

Web-based meta-documents provide us a robust and unobtrusive mechanism to re-orient Web documents towards specific user groups and attain tangible goals. Walden's Paths incorporates features that allow viewers of paths to customize their interaction to suit their preferences and immediate context. Managed collection of meta-documents along with the customized interface form a basis for providing reliable library services over distributed dynamic digital collections that can change in an unpredictable manner.

Acknowledgments

This material is based upon work that was supported by the National Science Foundation under

Grant Nos. IIS-9812040 and DUE-0085798. This paper expands and updates material previously presented in (11), (15), and (19).

References

BUSH, V. As We May Think. The Atlantic Monthly. August 1945, 101-108.

28 HALASZ, F., MORAN, T. AND TRIGG, R. Notecards in a Nutshell. In: Proceedings of the

ACM CHI+GI Conference 1987. Toronto, Ontario, April 1987, 45-52.

TRIGG, R., "Guided Tours and Tabletops: Tools for Communicating in a Hypertext

Environment", ACM Transactions on Office Information Systems. 6(4), October 1988, 398-414.

ZELLWEGER, P.T. Active paths through multimedia documents. In Document Manipulation and Typography (Proceedings of the International Conference on Electronic Publishing,

Document Manipulation, and Typography. Nice France, April 20--22, 1988, 19-34.

ZELLWEGER, P., Scripted Documents: A Hypertext Path Mechanism. In Proceedings of

Hypertext '89. New York, November 1989, 1-26.

SCHATZ, B. AND CHEN, S. Building Large-Scale Digital Libraries. IEEE Computer 29(5),

1996, 739-748.

FURUTA, R., SHIPMAN, F., MARSHALL, C., BRENNER, D. AND HSIEH, H. Hypertext

Paths and the World-Wide Web: Experiences with Walden's Paths. In: Proceedings of Hypertext

'97: the Eighth ACM Conference on Hypertext. Southampton, U.K., April 1997,167-176.

SHIPMAN, F., MARSHALL, C., FURUTA, R., BRENNER, D., HSIEH, H. AND KUMAR, V.,

Creating Educational Guided Paths over the World-Wide Web. In Educational

Telecommunications, 1996: Proceedings of ED-TELECOM 96, June 1996, 326-331.

29 SHIPMAN, F., FURUTA, R., BRENNER, D., CHUNG, C. AND HSIEH, H. Using Paths in the

Classroom: Experiences and Adaptations. In Proceedings of Hypertext '98. Pittsburgh, PA, June

1998, pp. 267-276.

SHIPMAN, F. M., FURUTA, R., HSIEH, H., FRANCISCO-REVILLA, L., KARADKAR, U.,

RELE, A., SHENOY, G. V., BRENNER, D. A.. Using the Internet in the Classroom: Variety in the Use of Walden's Paths. In: Proceedings of ED-MEDIA 99: World Conference on

Educational Multimedia, Hypermedia, and Telecommunications. June 1999, 335-340.

FURUTA, R., SHIPMAN, F., FRANCISCO-REVILLA, L., HSIEH, H., KARADKAR, U. AND

HU, S. Ephemeral Paths on the WWW: The Walden's Paths Lightweight Path Mechanism. In:

Proceedings of WebNet (1) 1999. Toronto, Ontario, October 24-30, 1999, 409-414.

KARADKAR, U., FRANCISCO-REVILLA, L., FURUTA, R., HSIEH, H. AND SHIPMAN, F.

Evolution of the Walden's Paths Authoring Tools. In: Proceedings of WebNet 2000. San

Antonio, TX, Oct. 30 - Nov 4, 2000, 299-304.

My Yahoo. http://my.yahoo.com/ (accessed July 2001).

Sign in for Personalized Recommendations, http://www.amazon.com/ (accessed July 2001).

30 KARADKAR, U., FRANCISCO-REVILLA, L., FURUTA, R., SHIPMAN, F. AND ARORA,

A. Supporting Multilingual Paths over the WWW. In: Proceedings of the Third International

Conference on Asian Digital Libraries. Seoul, Korea, 6-8 December 2000, 177-184.

DEL GALDO, E. AND NIELSEN, J. (EDITORS) International User Interfaces. New York:

John Wiley & Sons Inc., 1996.

KOKKOTS, S. AND SPYROPOULOS, C.D. An architecture for designing internationalized software. In: Proceedings of Software Technology and Engineering Practice, Eighth IEEE

International Workshop on incorporating Computer Aided Software Engineering, Athens,

Greece, 1997,13-21.

RUSSO, P. AND BOOR, S. How Fluent is Your Interface? Designing for International Users. In

Proceedings of INTERCHI '93. Amsterdam, The Netherlands, 1993, pp. 342-347.

KARADKAR, U., FRANCISCO-REVILLA, L., FURUTA, R., AND SHIPMAN, F. Structuring

Access to a Dynamic Collection of Digital Documents: The Walden's Paths Virtual Directories.

In: Proceedings of Digital Documents and Electronic Publishing 2000. Munich, Germany,

September 2000.

GIFFORD, D., JOUVELOT, P., SHELDON, P. AND O'TOOLE, J. Semantic File Systems. In:

Proceedings of the thirteenth ACM Symposium on Operating Systems Principles. Pacific Grove,

CA, October 1991, 16-25.

31 HARDY, D., AND SCHWARTZ, M. Customized Information Extraction as a Basis for

Resource Discovery. In: ACM Transactions on Computer Systems, 14(2), 1996, 171 - 199.

GOPAL, B. AND MANBER, U. Integrating Content based Access Mechanisms with

Hierarchical File Systems. In: Proceedings of the third symposium on Operating systems design and implementation. New Orleans, LA, February 22-25, 1999, 265-278.

MARSHALL, C. Making Metadata: A Study of Metadata Creation for a Mixed Physical-digital

Collection. In: Proceedings of Digital Libraries '98. Pittsburgh, USA, 1998, 162-171.

LEVY, D. AND MARSHALL, C. Going Digital: A Look At Assumptions Underlying Digital

Libraries. Communications of the ACM, 38(4), 1995, pp. 77-84.

WIEDERHOLD, G. Digital libraries, value, and productivity. Communications of the ACM,

38(4), 1995, 85-96.

Yahoo! - Frequently Asked Questions. http://docs.yahoo.com/info/faq/faq.html (accessed August

2001).

FRANCISCO-REVILLA, L., SHIPMAN, F., FURUTA, R., KARADKAR, U., AND ARORA,

A. Managing Change on the Web. In: Proceedings of First ACM/IEEE-CS Joint Conference on

Digital Libraries, Roanoke, VA, June 24-28, 2001, 67-76.

32 FRANCISCO-REVILLA, L., SHIPMAN, F., FURUTA, R., KARADKAR, U., AND ARORA,

A. Perception of Content, Structure, and Presentation Changes in Web-based Hypertext. To appear in: Proceedings of Hypertext 2001. Aarhus, Denmark, August 14-18, 2001.

BERST, J. Why Personalization is the Net's Next Big Thing, http://www.zdnet.com/anchordesk/story/story_1977.html (accessed August 2001).

SHIPMAN, F., FURUTA, R., BRENNER, D., CHUNG, C., AND HSIEH, H., Guided Paths through Web-Based Collections: Design, Experiences, and Adaptations. Journal of the American

Society of Information Sciences (JASIS), 51(3), 2000, 260-272.

Number of Words excluding the references: 5,681

33