Hypermedia Applications1
Schools 1
Museums 2
Industry 2
Diagnostics 2
Replacing Expert Systems 3
Knowledge Acquisition 4
Office Information Systems 5
Hypermedia Applications
- Schools
Several large and potentially dramatic hypertext projects support instruction. Intermedia , at the Institute for Research in Information and Scholarship (IRIS) at Brown University, is the largest project, having developed courses in literature and biology. A videotape of the system is available from Annenberg (Call toll free 1-800-LEARNER). Project Perseus at Harvard University is developing hypermedia in Greek literature and the arts.On the one hand all extant classical Greek literature will be made available on a CD-Rom, and as another part of the project an elaborate hypertext system will be created in HyperCard, on the nature of the Hero in Greek Literature. This part of the project will create instruction in classical greek, put on
line maps and descriptions of Greek artifacts, and create a truly effective instructional environment.of these projects are non-linear and divergent thinking processes. However, the learning effects seem more pronounced for those involved in developing the text than for those merely browsing the text.
Projects with HyperCard are becoming too numerous to catalog completely. The Macdonnell foundation is supporting an extensive series of efforts. One of the most advanced of these is the American Culture in Context: Enrichment for Secondary Schools (ACCESS) led by Kathryn T. Spoehr of Brown, and in cooperation with IRIS. This project is designed to enrich high school courses in American Literature and History with a HyperCard corpus of text, picture, auditory, and video materials organized by teachers in an interdisciplinary collaboration. The process of creating these materials will be studied in context both ethnographically and to discover basic cognitive mechanisms and principles. Finally, the project will study how the new materials and technology affect work habits and classroom participation. If there are changes, how do these changes spread through the school among both students and teachers?
Before the materials were created, an extensive study was carried out of the structure of knowledge among teachers ( experts ) and students (novices) of American history and literature. Standard techniques of knowledge structuring were used: similarity ratings, typicality ratings of exemplars of important concepts, and a reading and recall task. As usually found in these studies, experts' knowledge was arranged hierarchically with greater depth and business of the knowledge trees. Experts' grouping were based on abstract criteria with greater embedding; while novices' structures were based on physical characteristics and were much flatter in depth. This work is continuing and promises to show how technology can be better used to alter education both inside and outside the classroom.
Katz, D. B. Conceptual structure and the growth of expertise in American history. Department of Cognitive and Linguistic Sciences, Brown University, Working Paper, 1989
Zeitz, C. M./ Expert-novice differences in memory and analysis of literary and non-literary texts. Doctoral Dissertation, Brown University, 1990.
GOVA <
In large scale software engineering projects, managing the information can be difficult. So, Tektronics has developed a hypertext system named CASE to help them solve that problem. In CASE, related nodes contain requirements and specifications, design notes, implementation notes, source and object code, test results, and user documentation. So, while working on any part of the software, designers have immediate access through their relational database to the information about the part they are working on. As ideas occur, the designer jumps to implementation notes. This application resembles an online documentation system. Rather than providing documentation for a completed system, software engineering tools provide information and code that help designers build the system. It is a more constructive application.
One of the more interesting interface modifications for this diagnostic system is the replacement of the cursor/mouse interface with one for use with a touch screen. Service mechanics are generally unfamiliar with computers, and there is no scope in the service bay environment for a mouse and keyboard.The constraints of the touch screen display made it necessary to confine display to a single document at a time, to widen the scroll bars, and to replace menu commands with action buttons.
One of the most intriguing reports to come out of Hypertext '89 was a short paper tucked away into a symposium. It did not receive the attention it seems to deserve, and did not even make it into the proceedings of the meeting. But it has a practical implication that is truly astonishing. Given all the hype and ballyhoo over the past decade about expert systems and their role in the future, this simple paper implies very straightforwardly that for many purposes hypertext systems are easier to develop and maintain, and just as effective.
The work began as a diagnostic system for complex computer datalink services. Because expert systems had been widely used for similar purposes before (e.g. Clancey, 1986). A datalink is a telecommunication connection between central computers and world-wide applications. The connection is made over many different hardware, communication protocols, line speeds, and access methods. If the datalink fails, determining the source of failure involves a complex abductive and deductive process that needs to create multiple hypotheses and develop criteria for establishing or ruling them out.
The original expert system developed for this purpose, to run automatically whenever invoked by the description of a particular failure, required a complex expert system shell and specialized Lisp functions to develop. Over a period of more than 5 months two people created a system containing 250 rules, 160 parameters, and 30 specialized Lisp functions. It works by asking questions of an operator who is in contact with a customer (complaining of a problem) over the telephone. The system takes the operator's answers and uses standard inferencing techniques to suggest what the problem is and its potential solution.
After the expert system had been developed, the developers returned to discover that they no longer remembered how it worked, nor understood its overall flow. The underlying structure was spread out over many rules and parameters, and was very difficult to understand. It occurred to them that KNOW, an in-house hypertext system without an inferencing or rule representation powers, might be able to help them derive a consistent overview of the diagnostic flow. Simple diagrams were developed in KNOW to capture the flow of reasoning in the expert system. Each component of the diagram was a hypertext link selectable by the user. Selecting the link can bring up a component description, a list of procedures to follow to determine its failure state, or a subsequent step in the diagnosis. Meanwhile, the operators can still have the browser overview available to suggest alternatives or to keep them aware of where in the diagnostic procedure they are. This profound navigational tool helps orient them within the complex procedures.
Peper's experience is that maintaining and updating the knowledge base within the hypertext format is considerably simpler than with the expert system. If a problem is found, the operators or developers can fix it immediately and directly in the text or flow diagram. This contrasts sharply with the expert system where the "consultation" must be ended; the development environment must be entered and the faulty rule, parameter , or Lisp function altered, with unpredictable results; then the expert system must be run again to see if the problem was indeed fixed. However, at that point, other unseen difficulties may still be in store for the unwary developer. Because of the complex ways in which rule interpreters work, there may be side effects of the changes that show up only later in interaction with other rules. This difficulty is completely avoided by using hypertext.
In the most important aspect, the speed and accuracy of actually finding problems and resolving them, there were no great differences between the expert system and the hypertext application: the hypertext system led to more solutions, but it required more outside assistance and more time. But in all other important ways, the hypertext system was judged superior.
Understanding: Operators were much more confident of their solutions, and rated the expert system more helpful, perhaps because they could understand the reasoning of the system and follow its logic.
Preference: Operators preferred the hypertext system by a wide margin: 56 % versus 25 % for the expert system, and 19 % with no preference.
Overall, these findings have general implications for the increased usefulness of hypertext systems over expert systems. Specifically, diagnostic systems should be examined for the possibility of improving their usefulness with a hypertext format. However, the kinds of advantages shown here for hypertext may well hold for many complex reasoning tasks where a clear network or hierarchical sequential structure is possible: expert systems for scheduling, for filling out forms, for advising and consulting, for management, and for bureaucrats. Obviously, the role of hypertext for instruction, to capture the intelligent tutoring of a creative and experienced teacher falls in this category too.
Foreign Language Instruction
Edited volumes are being prepared by Jonassen to cover a host of topics.
<
- Museums
Diagnostics
An interesting example is in a service bay diagnostics system built for Ford Motor Corporation. The implementation is for a Hewlett Packard Vectra computer. The system provides for a connection to an on-board diagnostic computer in a vehicle. Expert system technology provides diagnostic support, and is linked to the hypertext documentation, largely held on CD-ROM. This information is supplemented from a remote database connected to the diagnostic system over a dial-up circuit.The expert system contains embedded references to CD-ROM based hypertext documents, which can also be accessed through tables of contents as described earlier. If the user wishes to read documents held on the database, they are converted to hypertext format at runtime by routines similar to those used for batch process text file conversion.The interface to the three subsystems is designed to be transparent. A simple combination of pointing and scrolling, allows the user to select a mode and a subset of documents. Regardless of the source format, documents are displayed in a consistent manner.
Replacing Expert Systems
This table summarizes their findings:
Knowledge Acquisition: Experts themselves can enter text in the appropriate places, as comments or instructions. There is no need to learn about rule languages or the complexities of rule interpretation by the computer.
- Office Information Systems
The find of 1990!
There is a utility out there called LZEXE.EXE which
compresses .exe files to a fraction of their former sizes and then runs them
ON THE FLY! The files run significantly to exponentially faster.
The funny part is that a kid named Fabrice Bellard wrote it and when he
tried to sell it to American companies they pooh-poohed the idea because the
documentation WAS WRITTEN IN FRENCH! ( Talk about ethnocentricity...)
Anyway, Fabrice copyrighted it and put in in the public domain so no one
could sell it. Thus it is available to anyone.
I have used it on Wordperfect ( runs about twice as fast ), and some real
old type IBM languages, like the original IBM LOGO. You have got to see that
turtle move! This is NOT your typical pkzip utility. This compresses .exe files
and runs them compressed. I predict this will be THE program of 1990!