©1995 Brian M. Winn
2.3 Note Taking
2.5 User Statistics
3.0 Future Directions
3.1 Enhancements to Current System
3.2 Beyond Current System
The Knowledge Encoding and Retrieval Tool (KERT) project is an attempt to resolve many of the deficiencies found in hypermedia systems. KERT is composed of an underlying hypermedia structure coupled with a comprehensive set of navigational and learning tools to more fully realize the power of hypermedia for thinking and learning. It is both a multimedia developers tool and an end-user environment.
A KERT knowledge web contains a large collection of "documents," each of which integrates multimedia content elements (text, sound, graphics, digital video, and interactive experiences). Guided discovery learning and reduced navigational disorientation are achieved by providing the user with a series of paths through the knowledge web. A path is a linear series of documents that represents a focused, contextualized tour through part of the nonlinear web of documents. "Contextual glue," in the form of comments and questions, is attached to each document in a path to clarify how the document fits into the context of the path, to provide a smooth flow of information between documents, and to provide a means of testing the user's comprehension. A document can be contained in multiple paths, hence a user can view the same document from multiple perspectives. Associational hyperlinks within the documents provide users with opportunities to explore their own interests that may be tangential to a path, or to seek expanded information on important concepts.
A log of experiences is automatically constructed as the user navigates through the hyperweb. The log can be used to review or return to any document that has been viewed. Users can construct their own paths using the log of experiences and their own contextual glue. These paths can then be distributed to and explored by others.
Hypermedia, also known as interactive multimedia, has been heralded as the next-generation learning tool . Hypermedia systems have the ability to incorporate large bodies of information with multiple forms of media representation (text, graphics, audio, video, and interactive experience.) The real power of hypermedia as a learning tool lies not only in the wealth of information that any one system can contain but also in the potential of the system to give the user access to the "right" information at the "right" time and the proficiency of the system to present the information to the user in a representation that motivates the user and promotes learning . Unfortu-nately, these two capabilities are often missing entirely and have yet to be fully realized in hypermedia systems today.
The Knowledge Encoding and Retrieval Tool (KERT) project is an attempt to address these chal-lenges in hypermedia system design. Optimal ways of encoding knowledge into a hypermedia system (structuring content) and optimal ways of retrieving knowledge from the system (search-ing, navigation, and interactive experiences) are focal points of the research.
The knowledge encoding and retrieval process, referred to here, is not the artificial intelligence problem of encoding knowledge for use by the computer. Rather, it is the human challenge of encoding and retrieving knowledge. How can humans use hypermedia to optimize the encoding of their knowledge for retrieval by other humans?
The KERT research is progressing in conjunction with the production of several domain specific multimedia projects in the Communication Technology Lab. Techniques developed in the KERT research will be incorporated into many of these projects and studied as they are used by teachers and students to gage their effectiveness.
This section defines the primary elements of KERT: the Document, the building block of the knowledge domain web; the Path, a method for guided discovery learning through the web; the Note tool, a means for users to augment the information they encounter; and the Journal, a tool for recording and organizing learning experiences in the hypermedia system.
The KERT document is similar to documents found in typical hypermedia systems today. However, KERT documents are built on an underlining framework that emphasizes the structur-ing of content in a form the user can easily understand. Further, KERT documents can be constructed using proven, industry standard authoring tools.
Structuring of Content
Typically, knowledge domain experts create a set of KERT hypermedia documents to be included in the system. Each document represents a conceptually associated chunk of information. These chunks of information are in the form of various media types, including text, graphics, sound, animation, digital video, and interactive experiences, organized into an end-user representation.
KERT documents are structured into screen-sized pages. Paged documents were chosen over scrolling documents and over multiple overlapping windows to maintain full control over screen layout and media integration. Studies have shown that spatial layout promotes recall by providing the user with a "visual mnemonic" for storing and retrieving the content of the document . Kyoichi Arai et al. claim that with scrolling fields, users often has a difficult time recalling where previously viewed information was located and further, they even have a difficult time recogniz-ing previously view information because it often does not appear in the same screen location . Navigating between screen-sized pages reduces confusion about the logical structure of the docu-ment by making it easy for the user to determine where he or she is in the document and how far they have to go (i.e., "page 2 of 10").
However, following a strict pagination structure can be limiting to a multimedia designer in that it does not allow for nonlinearity within the document's contents. For example, a designer may wish to have an interactive map with several points of interest on it. When a user clicks on one of these points, information about that point is displayed. Users have the option of clicking on as many or as few of the points as they wishes. If this map were forced into a linear, paginated structure, page one would have the map and successive pages would contain the information about each point. This forces users to navigate through potentially large volumes of information which is not of interest to them rather than allowing a direct jump to the next logical content section (See Figure 1.)
To alleviate such problems, without losing the benefits of paged navigation through documents, KERT defines a page as a logical section. This may be a single screen, or it may be a collection of screens tied together with a menu or an interactive graphic. Each screen is called a location in KERT. So, in the map example, the map would be the first location in the page, the one that is displayed when the user turns to that page. The information about the points of interest would be in other locations in that page. These locations would be displayed only if the user clicked its corresponding hot point on the map. The next page button would display the next logical section of the document (See Figure 2.)
The design of the pages in the document is left up to the multimedia designer. However, it is suggested that each document contain, as its first page, a document overview. The function of the overview is to summarize the key document concepts. In addition, an overview can inform the user of which sections have visited since KERT keeps track of this data. This allows users to decide quickly if they would like to explore the document. In addition it allows users to determine if they have missed information contained in the document on a previous viewing.
Constructing the Document
Rather than creating an entirely new tool for the construction of multimedia documents in KERT, document authoring is built around the multimedia authoring system, Macromedia Director. Director was chosen over a proprietary system or other commercial systems for the following reasons:
´ Director is the premiere cross-platform multimedia development tool used on the Macintosh and Windows based machines. A version is also available for the 3DO multimedia box, with future versions in development. KERT documents created on one Director platform can be easily ported to other platforms, often with minimal rewriting of code.
´ It is a proven development environment that has been used for nearly ten years.
There is a wide base of supporting tools available for Director that could be used to help develop KERT documents. Further, there is a wide base of potential KERT multimedia designers.
´ It provides access to multiple media types.
A typical KERT document will include text, still images, animations, digital video, and sound. However, by using Director, KERT documents can also contain interactive media elements. For example, a document may contain an interactive "Food Chain" simulator. This simulator would allow the user to analyze and modify the interrelations of species in a food chain and analyze the outcome.
´ It allows for the integration of media elements:
Print media (i.e., magazines, books, brochures) have evolved sophisticated techniques for integrating text and images, aimed at aesthetics, readability, and impact . Hypermedia systems today tend not to integrate media elements very well. In most World Wide Web browsers, for example, several types of image files, digital movies, and sound get downloaded into their own "helper applications," often opening a separate window rather than appearing in a place of their own, integrated into the document. By using Director, KERT documents can have media elements which are placed strategically on their document pages to promote clear communication of meaning and media integration.
´ It allows programs to link to media element files, rather than embed them.
Separating the document structure from the media files is advantageous because the document and its media can be modified independently without affecting the other. More important, this separation allows multiple documents to share the same media without taking up extra space caused by duplicating media content. For example, a segment of digital video showing the Genip Tree, found in large numbers in secondary growth tropical forests, could be placed in a document about ecological succession. This video segment could also be included in a docu-ment about tropical rain forest ecology. Each document would include the same digital video, however, the video would be displayed in different ways, with differing captions and possibly different controls over the video. Re-using media elements in various contexts has the educa-tional advantage of helping students recognize the interrelatedness of a knowledge domain and encourages them to integrate concepts .
´ It contains a simple, yet powerful design metaphor. A novice, non-programmer, could learn to use Director without much difficulty. Yet, it has the power and expandability needed for com-plex programming tasks.
´ Existing programs written in Director can be adapted to work with KERT with a minimal effort.
Once there are documents in the KERT system, a variety of paths through subsets of the docu-ments can be created. Paths in hypertext have been proposed and implemented in different ways . In its most rudimentary sense, a path is a linear list of documents chosen from a web of inter-linked documents. Our implementation of paths is a hybrid of previ-ous path implementations.
Navigating a Path
The path is a navigational tool that can complement existing hypermedia systems. Paths resolve several deficiencies found in typical hypermedia systems without sacrificing the benefit of being able to explore nonlinear hypermedia links. They are elegant in their simplicity.
A path can be thought of as a focussed, guided tour through a subset of the full web of documents. Users can traverse a path linearly in the suggested order from document to document, jump around to the documents in the path in a nonlinear order of their choice, or move off the path through various hyperlinks.
A major problem in a totally nonlinear hyperweb is that users do not have a structure to define their exploration through the web. This lack of a structure results in navigational disorientation, the loss of one's way . A path provides a basic structure to users. It defines suggested nodes to visit in the web while still allowing users to explore and expound upon their own interests via hyperlinks. Therefore, the nonlinear capability of a hyperweb is not sacrificed; rather it is comple-mented. Users are never lost because they can always return to the path.
It has been found in previous hypertext systems that when users encounters navigational prob-lems, they often return to the starting point and begin their search again . The path provides a linear list of starting points. If users travel off the path, they can return to the most recently visited document in the path rather than go back to the beginning of the path. This reduces the amount of navigation users have to do to get back to where they were before following hyperlinks off the path.
In KERT, since a path represents a complete linear experience, conceptual disorientation is reduced . That is, users have a conceptual view of where they are, where they are going, and where they have been (See Figure 3.)
Some hypertext researchers suggest the use of visual maps or graphical browsers to show users the structure of the knowledge web and the user's current location in that knowledge web. Others, such as Dillon et al., argue that these maps do not help with user orientation because "...for richly interconnected material of a reasonable size and complexity, it is not possible to include everything in a single browser without the problem of presenting 'visual spaghetti' to the user. In such a case, it is necessary to represent the structure in terms of levels of browsers, and at this point there is a danger that the user gets lost in the navigational support system!". Each path in KERT represents one particular, focussed view of interconnections through the KERT hyperweb (as explained in "Promoting Flow of Information" section below). Therefore, there is no single, mappable visual representation of the structure of a KERT hyperweb.
Reducing Cognitive Load
Paths can also greatly reduce the cognitive load associated with the handling of link choices. Conklin asserts, "The path facility relieves the hypertext reader from having to make an n-way decision at each link; rather, the reader is provided a default pathway through the network (or part of the network), and can simply read the material in the suggested order as if he were reading a linear document." Zellweger notes that cognitive overhead is diminished because the path either makes or narrows users' choices .
Promoting Flow of Information
A path, in our implementation, represents a contextual view of the information. That is, the encap-sulation of documents within a path creates a context in which to view the documents. For exam-ple, one path in the environmental science domain could be a path about "Fresh Water Resource Management." This path might include documents about ground water and precipitation, desalin-ization factories, rainwater cistern operation and regulations, irrigation for agriculture, water pollution issues, and geological determinants of ground water availability. Each of the documents, when viewed through this path, will be viewed in the context of "Fresh Water Resource Manage-ment".
Although these documents are included in one path, that path does not own them exclusively. For example, a document about "Ground Water and Precipitation" may be included in the "Fresh Water Resource Management" path but it also is contained in the "Geology of Island Formation," "Biomes of the Islands," "Impacts of Deforestation," and "Sustainable Agriculture" paths. The document appears in a different context within each path. The document is constructed in a way that is conducive to informing multiple contexts. This can lead to a "jerky style" in which there is little or no flow between one document and the next . An encyclopedia is an example of a collection of documents which lacks flow between documents. Our paths contain contextual glue to bond the documents in a path together to show interrelationships smoothly and comprehen-sively.
Contextual glue is path-specific information associated with a document. It is created by the author of a path. (Multimedia designers create documents in Director. Content experts can author paths through the documents without needing to know Director or multimedia design.) Contextual glue changes, depending on which path is being followed to access a document. There are three forms of contextual glue. The first form, called External Glue, is contained within the overview screen of the document. Here, contextual glue explains why the document is being displayed (i.e., "how it fits into the big picture of the path's context"). The information, in the form of text anno-tations, is presented to the user only when the document is accessed through that path. Thus, document overviews will be different depending on which path the user is on when they access that document. The second form of contextual glue, called Internal Glue, appears inside of the document and includes commentary explaining what in the document is important in the particu-lar context. This information is conveyed through text comments that are displayed at each loca-tion within a document. The third form of contextual glue comes in the form of Challenge Questions. These questions, like Internal Glue, are attached to locations within a document. chal-lenge questions probe the users' understanding of the material in the context of the path. A docu-ment not accessed through a path will have no contextual glue.
Providing Guided Discovery Learning
Guided paths through documents enhance the learning experience for users. When children (or adults) visit a museum, they tend to move from display to display, missing much of the richness which could be conveyed by a docent who skillfully weaves perspectives, commonalities and differences, and personal stories about the artifacts. We want to provide perspectives and concep-tual frameworks to help learners know what to look for and get the most out of the content.
At the same time, we do not want to limit the user's freedom to access subject matter and construct their own interpretations. We do not want to force them into a mode of passively watch-ing. In Literacy Machines, Ted Nelson describes the teacher as a feudal lord, controlling the access to the subject matter under his or her own personal view . KERT allows for a diversity of perspectives and paths through a rich web of content which can best be approached and under-stood in many different ways. In addition, interactive learning experiences can be embedded within the linear, guided structure to provide motivation and active participation (see "Document" section above). Further, users can leave the paths to follow their specific interests through hyper-links and searching tools.
From the standpoint of an instructor, the challenge questions provide a means to asses the users' understanding of the material. As the user travels through the system, he or she is notified that a document location contains a question(s). The user can then open the Question and Answer tool to read and/or answer this question (See Figure 4.)
As previously mentioned, a single document can be viewed in more than one way, depending on the context or path it is viewed in. Therefore, when a user finds a document that is particularly interesting, he or she can ask what other paths contain that document (See Figure 5.) The user can then switch to one of these paths to view the document in a different context (i.e., with different contextual glue). At this point, the user may decide to follow this new path and abandon the origi-nal. Following a new path may provide more information or a different perspective on a topic the user finds interesting. The user may decide to return to the original path at a later time. Returning to previous paths and documents is facilitated through the Journal Log (as described in "The Jour-nal" section below).
Types of Paths and Picking Paths
Depending on the knowledge domain, a KERT system could contain any number of paths of vary-ing types. In addition, there can be different ways of choosing a path depending on the type of path. For example:
´ Expert Paths - These paths are explicitly guided by a human with a particular perspective. The path could include the expert's personal stories and perspectives on the material. It would be interesting to have multiple experts cover similar topics to see how their paths vary, both in what is in the path and in how the experts use contextual glue to tie the documents in the path together. The user could choose an expert path from a screen that gives an overview of each expert and the paths that they created.
´ Concept Paths - These paths are implicitly guided (with an anonymous rather than personified author) and are used to convey a certain concept. For example, there could be a path on "Biodiversity" in the environmental science domain. A user could select concept paths from a diagram showing the interrelationship of multiple concepts, such as in a taxonomy, or from a concept index.
´ Spatial Paths - In many knowledge domains, there may be natural paths through documents based on real world spatial localization. For example, a set of island biomes could be dis-played on a visual map of the island. The path corresponding to each biome could be selected from the visual map (See Figure 6.)
´ Textbook Paths - These paths represent the type of information found in a chapter or section within a chapter of a textbook. A user could select textbook paths from a table of contents, similar to those found in paper textbooks.
These four methods for picking different types of paths, called Path Pickers, are available to the user at anytime. Path Pickers are a special type of KERT document whose sole purpose is to inform users about the available paths and allow them to follow one of these paths.
As discussed previously, it is also possible to pick paths by accessing the path tool. This tool allows the user to find all the paths that contain a particular document. The path tool can access all paths in the KERT hyperweb (See Figure 7.)
Paths are stored separately from the hyperweb document database. Therefore, paths can be added and removed from a system. A publisher could offer sets of addition paths to consumers who have already purchased the hyperweb as an option or supplement. The user could also author his or her own paths (see section on Logs, Paths, and Lesson Plans below).
Alternatives to Path Navigation
The path is an excellent means of navigation for guided discovery learning. KERT also contains an alternative means of navigation called hyperlinks. Hyperlinks are user instantiate nonlinear jumps through the web of documents. The hyperlinks are primarily a tool for user-directed browsing of information.
There are two types of document hyperlinks in KERT; links that move from document to docu-ment, called Interdocument Links, and links that either display additional information within the current document or jump to a different location in the current document, called Intradocument Links. Interdocument Links are usually created by the designer of the system to represent associa-tional links. These links are usually represented within a document in underlined blue text and often prompted with a word phrase, such as "see also," followed by the document name. When users click on the text, they are presented with a Jump Previewer that contains the abstract of the linked document (See Figure 8.) At this point, the user can decide if he or she actually wants to leave the current document and go to the linked document. If the user decides to view the linked document, he or she can quickly return to the previous document via the Journal (See "The Jour-nal" section below.) Gray asserts that Interdocument Links are the primary cause of user disorien-tation. "Mistrust of the links as being the feature of the system that gets them off the topic or causes them to miss some information are manifestations of the disorientation that nonsequential links collabortively construct." It is our hope that this mistrust will be reduced through the use of the Jump Previewer and the Journal.
Gray also professes that users often feel lost because they cannot tell what they have or have not explored . In KERT, if the linked document has previously been viewed by the user, the Jump Previewer will display a visited indicator flag. This flag shows users whether they have previously accessed all, some, or none of a documents contents. Therefore, the user is less likely to revisit a previously viewed document inadvertently . The visited indicator is also displayed in the Path Tool and Journal for documents that have previously been viewed.
The Intradocument Links do not move the user to a new document. They incorporate new infor-mation within the body of the document, jump to another location within the current document, or popup a window of information. There are three common uses of Intradocument Links. First, Intradocument Links are used to bring up definitions of certain words or phrases. Second, due to the limited screen real estate, media elements such as images, digital video, and interactive elements often have to be displayed on a document in a scaled format without all their interactive controls. Intradocument Links allow the user to toggle between the scaled media element and its full sized complement. Third, some documents contain a graphical interactive multimedia menu of topics in the document. The menu contains Intradocument Links that allow the user to jump to different locations within the document.
Both forms of hyperlinks can be utilized not only within the contents of a document, but also within a path's contextual glue. Therefore, a path author can add links to the web of documents that may not already exist.
Rather than having a simple text editor for recording all notes, KERT allows the user to add indi-vidual notes for each location in a document (See Figure 9.) By creating in-context notes, the link between the note and the content is explicit. Therefore, when the user is reviewing his or her notes at a later date, he or she can quickly, via a link in the note, jump back to the location in the docu-ment in which it was taken.
The Journal represents a comprehensive tool that automatically tracks and records user experi-ences through the hypermedia web. The heart of the Journal is the log, a historical list of the all documents visited within the hyperweb by the user. The journal also contains tools for reviewing the user's notes taken and challenge question encountered, as well as a tool for constructing a user-created path.
This feature automatically records each step the user travels through the web of documents. "Studies of Videotext systems report that backtracking is the most important feature for novice users so they can back out of a situation they are uncomfortable or unfamiliar with."  There-fore, the log can be used as a navigational tool to return to any document that has already been accessed.
Users can also use the log to review the experiences they have encountered. The user has the option of selecting the documents in the log and jumping back to them for further review. The document will be displayed within the context (in other words path) in which it was viewed. The user can also review, for any given document in the log, the notes that he or she took (See Figure 10.) A link is present in each note, that when clicked, will take the user to the associated document and location where the note was taken. Notes can be created or added to a selected document from within the Journal. These notes will be linked to the associated document as a whole, not to a specific location within the document. The user can also inspect the challenge questions, for any given document in the log, and edit the answers. As with the notes, a link is present in each question summary, that when clicked, will take the user to the associated docu-ment and location where the question was encountered.
Logs, Paths, and Lesson Plans
Users can employ the Path Builder Tool in the Journal to create their own paths (See Figure 11). Thus, a student or teacher can create his or her personal tour through the hyperweb, complete with "expert" glue, and then distribute this path to others. The idea of building a path through the hyperweb of documents is synonymous to performing a survey of literature in which the web of documents represents literacy in a given domain. A path through the hyperweb could also repre-sent a lesson plan, a list of suggested readings, or even the result of a homework assignment that can be created and turned in for class credit.
The creation of contextual glue is facilitated by providing the user with a glue editing tool (See Figure 12.) This tool can be accessed as the user follows the path he or she is constructing. The tool is not available as the user travels other paths. The building of contextual glue in the KERT system is no more difficult to perform that taking notes or answer questions.
A path designer may want to create a path that contains other paths. For example, an instructor may like to make a reading assignment path that includes various textbook paths. The ability to have paths contain other paths could lead to a tangled web of paths causing navigational disorien-tation to the user, diminishing the benefits of the path and reducing paths to just another form of hyperlinks. Paths containing other paths would also complicate the path tool interface. Therefore, a path in KERT cannot contain another path. If the path designer wants to present interrelated collections of paths, they can design customized Path Picker Documents (described in "Types of Paths and Picking Paths" section above) to package multiple paths together.
KERT monitors and tracks the actions of the user. These user statistics, which include the log of documents visited, the notes the user took, and the answers the user wrote can be stored to an external file. This file can be loaded back into KERT to restore the user statistics (see Figure 13.) This facility is of interest not only to a user that wishes to save and resume his or her session, but also to content designers, researchers, and instructors. For example, they may wish to collect the user statistic files and use them to help analyze the effectiveness of the tool in passing knowledge on to the user.
The construction of KERT, to date, has been primarily proof-of-concept prototyping. In its present state, KERT is a powerful multimedia developers tool and end-user environment. There are a few features, however, that we would like to incorporate in the current system before we consider it a complete prototype and move on to the second stage of development.
The future enhancements to the prototype system are both end-user tools; a note structuring tool and a document searching mechanism. The note structuring tool will increase the usefulness and power of the note taking facility of KERT. The searching mechanism will allow the user to quickly access specific documents or paths.
Note Structuring Tool
The question arose early on: "Why would one want to take notes within our system? Why not just use pencil and paper?" An obvious response is that the tool allows the user to create in-context notes linked to the content. Further, computerized notes can be easily manipulated, printed, or exported to a text file. They can even be turned in as homework assignments.
These capabilities are commendable, but the system should provide more. Taking Donald Norman's philosophy, computers are tools that should complement our capabilities and provide us with facilities to make us smarter .
Ted Nelson said, "The intrinsic structure is what the user should see in a document, not the obtru-sive conventions under which it is stored." These intrinsic structures are inherently goal-oriented. One user of the system will have a different goal he or she would like to achieve in browsing a KERT hyperweb than another user. If the user takes the time to find and identify the intrinsic structure as it relates to his or her goals, he or she will come away will a much better understanding of the material. If, however, we explicitly defined a structure for the users, it would most likely not fit in with their goals, and the structuring may even impede their understanding of the material.
By providing a tool that simplifies the process of structuring knowledge, we hope that users will be more willing to identify the intrinsic structure of the material (as it relates to their goals) and, thereby, learn the material better. Further, we do not want the users to feel that this was an exer-cise in futility that resulted in just a higher-level of understanding (although that should be enough of an incentive.) There needs to be a motivation for the user to take the time to structure the notes. The motivation comes in that once the notes are structured within the system, a study guide is created that the user can use as a study aid in a classroom, testing, or research situation.
In developing the note structuring tool, we hope that we can devise an approach to knowledge structuring that is familiar and comfortable to the user. Further, we want the process of knowledge structuring to complement the user's own internal (cognitive) processes.
Two knowledge structuring approaches were reviewed: concept maps and hierarchical structur-ing. A concept map, derived from the work on human associative memory by Collins and Quillian  and cognitive maps by Tolman , is a spatial organization of domain concepts, represented by nodes, and the relationships between these concepts, represented by links. Concept maps allow the user to organize their knowledge by clustering of nodes and define the interrelations of the knowledge by defining links between the nodes . Cognitive psychologists posit that humans constantly form these concept maps, or semantic networks, within their mind . However, studies have found that humans often have a difficult time defining concept maps . Further, a concept map, being primarily spatial in nature, can become overwhelming if the map contains too many nodes and links. Further, they do not lend themselves very well to the construction of a study guide. Hierarchical structuring is an abstract organizational technique in which domain concepts, represented by nodes, are arranged in a tree structure, represented by branches between nodes. A node can have multiple children, but may only have a single parent. Hierarchical struc-turing allows the user to organize his or her knowledge within a tree structure. The tree may contain some implicit interrelations between knowledge through the parent-child structuring. It is impossible, however to explicitly define interrelations between nodes in a tree, as done in concept maps, due to the rigid parent-child structure of the tree. Despite this shortcoming, hierarchical structuring is the most widely used technique for organizing information and it lends well to the process of constructing outlines. Therefore, the hierarchical structuring technique was chosen.
The note structuring tool will work in conjunction with an enhanced Journal log that shows note titles, as well as the documents. The user can simply drag notes from the log into the hierarchical structure in the note structuring view. The process of dragging and dropping notes was chosen because direct spatial manipulation of concepts seemed the most appropriate method for creating and refining hierarchical organizations. The tool will function like the graphical file management tools available on most computers. The note outline created by the user can be printed or saved as a text file.
The only method of locating information in the current KERT system comes is through user browsing. There is no facility to search for a specific document, path, or piece of information. We would like to begin by creating a simple keyword boolean search capability that allows the user to search for paths or documents. The result of a search for paths would be a Path Picker that displays which paths were found. The result of a search for documents would be a path that contains the found set of documents.
More advanced searching mechanisms may be investigated in the future. For example, the system could utilize the user's usage statistics to build a profile of his or her interests. This profile could then be used to perform an intelligent search whose results could be offered to the user in the form of an Intelligent Agent Path Picker or Path.
The second stage of KERT will be to implement a knowledge base into the system and see how hypermedia designers and content experts work with the system. We will then move on to user testing, the creation of additional developer tools, and a complete code rewrite of the KERT system to increase its power and flexibility.
The KERT prototype will be used to construct a multimedia learning program in a specific domain. Extensive user testing will be done with the program to gage the effectiveness of the KERT environment. KERT will also be tested and compared to other hypermedia systems. From these studies, we hope to prove the validity of the KERT. We also hope that it will aid in modify-ing KERT to be a more effective knowledge encoding and retrieval tool.
Web, OpenDoc, and Beyond
The real power of KERT will not be fully realized until it has at its disposal a massive amount of information. The best way to achieve this is to make a networked version of KERT. Rather than create an entirely new network paradigm, it would be much easier and more successful to adapt it to work with the current World Wide Web.
The final stage of KERT will be the recoding and breaking up of its modules into OpenDoc parts. OpenDoc is a cross-platform software architecture that enables the development of component software through the combining of software parts. By breaking KERT up into separate OpenDoc parts, it will gain easy access to the Internet by being able to interface with the upcoming Internet OpenDoc tools, such as Apple's Cyberdog. Cyberdog includes a collection of communication parts (HTTP, ftp, Gopher, and NNTP), file-translation parts (HTML, GIF, JPEG, AIFF), and inter-face parts (bookmarks, notebook). Selective KERT parts could then be used to augment the access to the Web. Users could employ just the KERT parts they desired. Further, they could increase the functionality of KERT by replacing some of its parts, such as the primitive text editor component used in the note taking facility, with more powerful parts, such as a fully-functional word process-ing component.
This manuscript outlines a comprehensive set of hypermedia tools for knowledge encoding and retrieval. These tools are at the foundation of the KERT project, under development at the Communication Technology Lab at Michigan State University. We describe a knowledge web comprised of multimedia documents linked navigationally and contextually by interrelated, expert-guided paths. Path picker interfaces provide perspectives and interrelationships among paths. The Journal Log tracks navigational history for review. In-context note taking allow users to add comments and observations throughout the knowledge web. Challenge questions allow instructors to determine the understanding of the users as they travel the paths of information and help users learn by applying their understanding to new situations. The path builder empowers users to transform their own experiences through the knowledge web into paths that they can share with others.
In general, these tools can operate independently of each other. However, taken together they are tightly coupled in both functionality and form, complementing, supporting, and enhancing each other in a unified realization of the potential of hypermedia as a tool for learning and knowledge discovery.
 Jonassen, David H. "Conceptual Frontiers in Hypermedia Environments for Learning." In Journal of Educational Multimedia and Hypermedia; v2 n4, 1993. pp. 331-335.
 Bower, G. (Ed.). The Psychology of Learning and Motivation. Orlando, FL: Academic Press. 1988.
 Norman, Donald. Things That Make Us Smart: Defending Human Attributes in the Age of the Machine. Readings, Mass.: Addison-Wesley Publishing Company. 1993.
 Norman, Donald. "Psychological Issues in Support of Multiple Activities." In Defending Human Attributes in the Age of the Machine CD-ROM. Voyager. 1993.
 Arai, Kyoichi; Yokoyama, Teruo; and Matsushita, Yutaka. "A Window System With Leafing Through Mode: BookWindow". In Proceedings of CHI, 1992 (Monterey, CA, May 3-May 7, 1992) ACM, New York. 1992. pp. 291-292.
 Tufte, E. R. The Visual Display of Quantitative Information. Cheshire, CT: Graphics Press. 1983.
 Tufte, E. R. Envisioning Information. Cheshire, CT: Graphics Press. 1990.
 Spiro, Rand and Jehng, Jihn-Chang. "Cognitive Flexibility and Hypertext: Theory and Technology for the Nonlinear and Multidimensional Traversal of Complex Subject Matter." In Cognition, Education, and Multimedia: Exploring Ideas in High Technology. Nix, Don and Spiro, Rand (Ed.). Hillsdale, New Jersery: Lawrence Erlbaum Associates Publishers. 1990. pp. 163-205.
 Damarin, Suzanne K. "School and Situated Knowledge: Travel or Tourism?" In Educational Technology; v33 n3, Mar 1993. pp. 27-32.
 Roth, Wolff Michael. "Bridging the Gap between School and Real Life: Toward an Integration of Science, Mathematics and Technology in the Context of Authentic Practice." In School Science-and-Mathematics; v92 n6, Oct 1992. pp. 307-317.
 Zellweger, Polle 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. J.C. van Vliet (ed.). Cambridge, Mass: Cambridge University Press. 1988. pp. 19-34.
 Trigg, Randall H. "Guided Tours and Tabletops: Tools for Communicating in a Hypertext Environment." In ACM Transactions on Office Information Systems, Vol. 6, No. 4, October 1988. pp. 398-414.
 Garzotto, Franca; Paolini, Paolo; and Schwabe, Daniel. "HDM - A Model-Based Approach to Hypertext Application Design." In ACM Transactions on Office Information Systems, Vol. 11, No. 1, January 1993. pp. 1-26.
 Marshall, C. C., and Irish, P. M. "Guided Tours and On-line Presenations: How Authors Make Existing Hypertext Intelligible to Readers." In Proceedings ACM Hypertext '89 (Pittsburgh, PA, Nov. 5-8, 1989). pp. 15-26.
 Gray, Susan H. Hypertext and the Technology of Conversation: Orderly Situational Choice. Greenwood Press, Westport, CT. 1993.
 Merlet, Jean-François. "Human Computer Interactions Through an Intelligent Hypermedia". In IEEE Proceedings on Man, Machines, and Cybernetics. 1993. 263-271.
 Carey, T. T.; Hunter, W. T.; and Lopez-Suarez, A. "Roles for Tables of Contents as Hypertext Overviews". In Human-Computer Interaction - INTERACT '90. D. Diaper et al. (Ed). Elsevier Science Publishers B.V. (North-Holland). pp. 581-586.
 Stanton, N. A. and Baber, C. "The Myth of Navigating in Hypertext: How a "bandwagon" has lost its course!". In Journal of Educational Multimedia and Hypermedia, Volume 3, Number 3, 1994. pp. 235- 150.
 Yankelovich, Nicole; Haan, Bernard J.; Meyrowitz, Norman K.; and Drucker, Steven M. "Intermedia: The Concept and the Construction of a Seamless Information Environment". In IEEE Computer. January 1988. pp. 81-96.
 Conklin, Jeff. "Hypertext: An Introduction and Survey". In IEEE Computer. September 1987. pp. 17-41.
 Dillon, Andrew; McKnight, Cliff; and Richardson, John. "Navigation in Hypertext: A Critical Review of the Concept". In Human-Computer Interaction - INTERACT '90. D. Diaper et al. (Ed). Elsevier Science Publishers B.V. (North-Holland). pp 587-592.
 Zellweger, P.T. "Scripted Documents: A Hypermedia Path Mechanism." In Hypertext '89, Pittsburgh 1989. pp. 1-14.
 Furnas, George W. And Zacks, Jeff. "Multitrees: Enriching and Reusing Hierarchical Structure". In Proceedings of CHI, 1994 (Boston, MA, USA, April 24-28, 1994) ACM, New York, 1994. pp. 330-336.
 Nelson, Theodor Holm. Literary Machines. Published by author, 1987.
 Nielson, Jakob. "The Art of Navigating Hypertext". In Communications of the ACM. Vol. 33, Number 3, March 1990. pp. 296-310.
 Luger, George F. and Stubblefield, William A. Artificial Intelligence: Structures and Strategies for Complex Problem Solving, Second Edition. Redwood City, CA: Benjamin/Cummings Publishing Company. 1993. pp 354-368, 694-703.
 Reader, Will and Hammond, Nick. "Computer-Based Tools to Support Learning From Hypertext: Concept Mapping Tools and Beyond". In Computers and Education. Vol. 22, No. 1/2, Jan-Feb. 1994. pp 99-106.
 Aschcraft, Mark. H. Human Memory and Cognition, Second Edition. Cleveland, OH: HarperCollins College Publishers. 1994. pp 254-307.
Back to Table of Contents