Pasi Häkkinen, Kari Suomela
Tampere University of
Technology
Department of Mathematics, Hypermedialaboratory
PO Box
692
FIN-33101 Tampere, Finland
e-mail:{pasi.hakkinen,
kari.suomela}@cc.tut.fi
The source material should be properly structured, e.g., in Word format the styles should be used consistently and comprehensively. Source material that is written carelessly or only layout in mind decreases the conversion quality. This is often the case with text that is originally intended for paper publication [1].
Assuming the source material is well-structured, our conversion process consists of four steps:
1. Use latex2html or rtf2html converter
2. Edit index file
3. Generate links
4. Make files available in the Web
Our work is supported by Academy of Finland, Finnish Ministry of Education and Tampere University of Technology.
The conversion process described works also in other domains.
IULM is the most popular of our Web courses. It is intended for students entering the university that do not have sufficient skills in mathematics. After taking the course, a student's mathematical knowledge should be at high school senior class level. The course has also been tested in one of the high schools in Tampere area.
The Matrix Algebra course teaches the basics of linear algebra. MA was our first trial in transforming material from paper format into Web.
BH was lectured for the first time during autumn 1996. We had no existing material, so everything was created from scratch. One of the three authors used LaTeX, another used Word and the third chose to write HTML directly. The course material was built for Web and is available only in that format - quite natural for a Hypermedia course.
All three courses have a selection of problems from previous examinations for practicing. For the same purpose, IULM also contains problems from high school final examinations and university entrance examinations from previous years.
IULM and MA were written with Microsoft Word. Details of the conversion process are described in [4]. A recently added feature of RTFtoHTML not described in that article is the automatic conversion of Word indexes. Using Word commands, the author marks definitions and other words in text that he wants to have in the index. The converter creates a separate file from the Word index that can then be used in the linking process.
Proper use of styles in Word documents is essential for a successful conversion process. Headers, body text, URLs and other logical entities should be marked using appropriate styles. A selection of such styles is collected in a Word template file that is used by the authors and maintainers of the courses.
As mentioned earlier, BH is written with three different tools. The HTML layout looks slightly different depending on the source format, which, of course, does disturb the reader to some extent.
The index files produced by the RTFtoHTML converter are processed by a filter that removes all non-essential information and leaves only the actual hot word and the link target. The resulting file, called the link file, consists of lines in the following format:
matrix -> http://machine.org/ma/defination/matrix.html
vector -> http://machine.org/chapter1/section2.html#vector
The leftmost column lists the hot words, the arrow is merely a visual aid for the human reader, and the rightmost column lists the corresponding link targets.
Finnish is an inflected language, for example, nouns have 16 cases. An additional complication is the changing of word stems. See, for example tapa, "manner", vs.
tavoissaan, "in their manners". This is a major problem for the automatic linker.
There are two basic solutions. First, we can list the hot words in every inflected form in the link file as a regular expressions. We have written a program called Linktool that searches the HTML text for occurrences of these word forms and links them to their targets.
A better solution is based on a program created by a Finnish company specializing in commercial language software. The program is called Morfo [7], and it is able to break down even the toughest Finnish words into their basic morphological components. This also gets around the problem of comparing word forms that look very similar or exactly the same, but nevertheless originate from totally different base words. Furthermore, Morfo can also analyze complete sentences grammatically, which further improves the quality of automatic linking.
Matlab [10] is a package for numerical calculations, especially for linear algebra. We have experimented with Matlab in the Matrix Algebra course using it the same way as Mathematica.
Java [11] is ubiquitous. The MA course contains an experimental Java applet that displays a matrix that the student is then requested to classify as singular, diagonal, symmetric etc. Java offers endless possibilities for simulations and animations. Java is run locally, which decreases the bandwidth requirements. Java is also platform-independent, which is a major asset in a heterogeneous environment like ours.
tasks a lot easier. A HTML page using SSI contains special commands disguised as comments. The server replaces the commands either with their output or with the contents of a file. The example below displays the current time, the contents of the file text.html and finally the last modification time stamp for the HTML file.
<html>
<title>SSI Example</title>
<body>
<!--#exec cmd="time" -->
<!--#include file="text.html" -->
<!--#flastmod -->
</body>
</html>
SSI increases server load, so it should be used with caution.
We have used SSI in building exercises. The exercises are modular: they contain the actual problem, links to prerequisite concepts, some hints and finally the solution. Different parts are visible according to the student's choice.
Figure 1. An exercise made with SSI.
The communication buttons are included in using SSI - this makes creating communication bars with uniform looks easy.
Images containing sharp contrasts, for example, black text on white background, are not very well suited for video compression.
Some browser plug-ins support pseudo-streams. This means that browser can start playing a Quicktime animation before it is completely loaded. Regular Quicktime movies must be processed by a special filter program to enable this support. A public domain program for such processing is, e.g., Internet Movie Tool program [15].
The quality of Quicktime animations is usually good, but this is in exchange for rather large video files. Quality reduction decreases file size, but reasonable space savings would make the animations useless. This is due to the low rates of Quicktime compression algorithms: they are inteded for use in Local Area Network (LAN) or faster environmants. To serve ISDN and modem users (28 kbit/s or more) we must find other methods for animation distribution.
Another way to distribute hypermedia courses is CD-ROM. Even large animations load fairly fast from a CD-ROM. We have produced a CD-ROM for IULM and MA. Connection to Web server is needed only for interactive exercises.
Figure 2. RealVideo animation embedded in an HTML page.
Stream-based animations offer a good solution for low-bandwidth connections. Such an animation is displayed at constant frame rate as it is downloaded. The animations are created by converting a AVI file into a stream format. This means that our Quicktime animations had to be converted into AVI using Adobe Premier. The resulting files are stored on a video server and sent over network from there.
At the moment there are products that are able to transmit real-time video over modem-speed connections. The quality of picture depends on the compression rate used. Using the best compression algorithms, the quality is sufficient for the animations we currently use in our Web courses.
We have tested the following packages: RealVideo [16], VivoActive [17], StreamWorks [18], VDOLive [19] and Vosaic [20]. They all use Video for Windows (AVI) as source format for compression; StreamWorks accepts only M-JPEG AVI files.
After testing we chose RealVideo and VivoActive. The decisive criterion was the compression method: both packages are capable of handling black text on white background with sufficient sharpness. Our animations contain a lot of such material, and this material is a major source of problems for most conversion algorithms. Using modem speeds, the image quality is reasonable but the frame rate is low (1 fps).
2. Microsoft Corp, RTF, ftp://ftp.primate.wisc.edu/pub/RTF
3. Chris Hector, RTFtoHTML converter, available at http://www.sunpack.com/RTF/
4. Eero Pantzar, Seppo Pohjolainen, Heli Ruokamo-Saari, Jarmo Viteli (eds.),Theoretical Foundations Applications Modern Learning Environments, Tampereen yliopisto, J[Sinvcircumflex]ljennepalvelu , Tampereen yliopiston tietokonekeskuksen julkaisuja n:o 2, 1996
5. Nikos Drakos, From Text to Hypertext: A Post-Hoc Rationalisation of LaTeX2HTML, First International Conference of the World-Wide-Web,Cern at 1994. Available at http://cbl.leeds.ac.uk/nikos/ tex2html/doc/latex2html/latex2html.html
6. W3C World Wide Web consortium , The Arena Web Browser, Available at http://www.w3.org/pub/WWW/Arena/beta-3
7. Kielikone, Morfo, http://www.kielikone.fi
8. Wolfram Research, Mathematica, http://www.wolfram.com
9. Kostadin Antchev, Markku Luhtalahti, Jari Multisilta, Seppo Pohjolainen, and Kari Suomela, A WWW Learning Environment for Mathematics, Fourth International World Wide Web Conference Proceedings, December 11-14, Boston, pages 251-258, 1995
10. MathWorks, Matlab, http://www.mathworks.com
11. Sun Microsystems, Java, http://www.sun.com/java
12. NCSA HTTPd Development Team, http://hoohoo.ncsa.uiuc.edu/
13. Apache HTTP Server Project, http://www.apache.org
14. Apple Computer, Quicktime, http://quicktime.apple.com/
15. Apple Computer, Quicktime Internet Movie Tool, http://quicktime.apple.com/dev/devweb.html
16. Progressive Networks, RealVideo , http://www.real.com
17. Vivo Software, VivoActive , http://www.vivo.com
18. Xingtech Technology, StreamWorks, http://www.xingtech.com
19 VDOnet, VDOlive , http://www.vdolive.com
20. Vosaic, Vosaic , http://www.vosaic.com