Publishing an Interactive Hypertext Book

Local time: Friday, 05-Dec-2008 12:14:28 EST
Last update at /special/conference/apwww95,Friday, 21-May-2004 09:47:18 EST

Publishing an Interactive Hypertext Book

Leslie M. Goldschlager
Department of Computer Science
Monash University, Clayton Campus
Victoria, Australia 3168

email: les@cs.monash.edu.au

Abstract
This paper describes a non-fiction book "Grandkids Australia" recently published on the World Wide Web. The paper discusses user-interface issues arising from hypermedia publication, and reports on experience gained over a five month period in monitoring actual Internet readership of the book.
A variety of navigational methods are described which facilitate readers actively mapping their own interactive trail through the book. Software is also provided which allows readers to add their own views and ideas to the book and thus gain a fuller interactive experience.
The paper concludes with lessons and suggestions for improvements which can be drawn from observing how the hypermedia book has been used on-line in practice.

Keywords
on-line books, usability, database software, hypermedia, hypertext, interactive publication

Introduction
This paper describes the rationale and technology behind the non-fiction work Grandkids Australia [Bergin 95] recently published on the World Wide Web. As the book calls on readers to be active citizens of Australia rather than passive spectators, the interactive nature of the Web presents unique opportunities for a hypertext rendition of the book to "practise what it preaches".
There is a considerable literature on design principles for hypermedia applications such as on-line books. It is beyond the scope of this paper to give a full overview of this large body of literature. Suffice it to say that an excellent annotated bibliography on "Graphic Design for the User Interface" [Lynch 95] is maintained on-line by Patrick J. Lynch of the Yale Center for Advanced Instructional Media at http://info.med.yale.edu/caim/Biblio_GUI.HTML. In addition, good collections of papers on the topic can be found in the recent Workshop on Hypermedia Design [Fraisse 95], and in the CACM Special Issue on Designing Hypermedia Applications [ACM 95] which is also published on the web at http://www.acm.org/siglink/.
Many on-line books have been published on the web. John Ockerbloom at Carnegie Mellon University maintains a web page called "The On-line Books Page" [Ockerbloom 95] at http://www.cs.cmu.edu/Web/books.html which currently lists more than 850 English works in various on-line formats. Unfortunately, many of these make little use of the established techniques for hypermedia design. Many on-line books are simply a single linear text file transcribed or scanned from the original paper form, or else are a set of text files, one per chapter, with a cover page containing one hypertext link per chapter. Some on-line books contain a richer set of hypertext links, but even these rarely pay sufficient attention to the exciting opportunities for different navigational paradigms which the medium allows.
The version of "Grandkids Australia" published on the Web allows readers to add their own views and comments, to share their own future visions, and to vote on issues raised by the reading material. More subtly, opportunities abound for allowing readers to break away from the normal linear structure of books published using traditional media. A well-designed hypertext book can include a wide variety of navigational methods which facilitate readers actively finding their own paths through the portions of the material which they find relevant.
This paper also discusses the software which facilitates the addition of new material by readers. As the data builds up in size, the software automatically creates indexes so that readers can more easily find their way around each other's contributions. The paper discusses the techniques used to extract relevant information from the book dynamically, particularly with the aim of creating short HTML responses in real time which gives reasonable response rates even to low bandwidth users.
Apart from usability and technology considerations, traditional book publication also entails appropriate graphic design principles. In the case of large collections of information represented in hypertext, such as hypertext books, graphic design issues are important in imparting a cohesive look and feel on the whole work, while empowering the reader to use the hypertext facilities which are offered.
The paper reports experience with the actual use of the "Grandkids Australia" book, and presents conclusions and improvements which can be drawn from the experience.

Overall Structure
The Grandkids Australia book divides naturally into chapters and sections within chapters. These were used as the natural structural elements of the hypertext design. However, care was exercised to avoid committing too early to the graphical look and feel of the individual web pages and to the placement and functionality of the navigational features. An all too frequent pitfall of web design is to allow those portions of the design to become too tightly integrated with the actual data. The result is to inhibit experimentation and discourage the implementation of major changes suggested by useability feedback. A large web site can have many hundreds of pages of information and if the graphical and navigational features are tightly built in to all these pages, it becomes a major task to carry out even minor alterations which need to be replicated on every page.
To avoid this difficulty and allow experimentation and flexible modifications, a database was created to store all the book sections without reference to look and feel or navigation. As on-line readers interact with the book and click on various navigational links, this database is accessed "on-the-fly". The data retrieved from the database is combined with the page templates which specify the look and semantics of the graphical and hypertext elements. The resulting web page, created by software as and when requested, is then sent to the on-line user in response to their request.
Apart from the flexibility of design afforded by this use of page templates, the technique also ensures a uniformity of layout and a consistency of navigational elements which would otherwise be difficult to maintain in practice across hundreds of individual pages on a web site.
The Grandkids Australia book has now been published on the World Wide Web for five months. Over this period, the site has received over 16,000 hits from over 3,400 different computers. This averages out to over 100 hits per day, or more than 4 hits per hour, 24 hours a day, 7 days a week. A "hit" is a request for either a page of text or for a picture. The actual number of requests for information is probably considerably higher owing to the practice of "caching" or storing web pages for reuse. This is common both at the browser level, and also on intermediate computers, particularly those of the Internet Service Providers.
Similarly, the number of readers is probably higher than the number of different computers, or "hosts", since some computers (and again particularly those belonging to the Internet Service Providers) frequently have multiple users. The following graphs show the distribution of hits among the top 20 computers, and then the next 100 computers which accessed the Grandkids Australia book. Ignoring hits involving pictures, the computers responsible for the most hits logged approximately 400 hits each. This dropped to approximately 40 hits for the 21st computer. From these figures, it is very likely that the top group of computers were each being accessed by quite a number of users.

Although the vast majority of accesses came from readers situated in Australia, about 20% of the accesses to Grandkids Australia were from a dozen other countries. It is natural that a book dealing with an Australian topic and available on an Australian server would be read mainly by people living in Australia. Nevertheless, the figures were probably skewed further towards Australia than would otherwise have been the case by the fact that Grandkids Australia was displayed at two "Multimedia Forums" hosted in Melbourne and Sydney by the Federal Government. The results by country are shown in the following table:
		  Accesses    Hosts

 Australia (.au)     82.9%    72.8%
 USA                 13.5%    20.8%
 Canada (.ca)         1.3%     2.2%
 New Zealand (.nz)    0.8%     1.3%
 Singapore (.sg)      0.8%     0.3%
 United Kingdom (.uk) 0.3%     0.4%
 Japan (.jp)          0.2%     1.2%
 Thailand (.th)       0.1%     0.1%
 Finland (.fi)        0.1%     0.1%
 Germany (.de)        0.1%     0.3%
 Belgium (.be)        0.1%     0.1%
 Malaysia (.my)       0.0%     0.1%
 Italy (.it)          0.0%     0.1%
It is interesting to note that whereas 73% of the computers were in Australia, these accounted for 83% of the accesses. The explanation is probably two-fold. Firstly, the book has a strongly Australian focus, being primarily a "call to citizenship" aimed at the Australian people. Therefore it would appear more likely that people from other countries, on reading a small segment of the book, might cease reading earlier than Australians. Secondly, the practicalities of bandwidth and speed on the Internet at present render it plausible that people from other countries would experience a slower response rate from an Australian site, and therefore be deterred from reading through a whole book on-line.
The next table shows a further categorisation of those 20% of computers which accessed Grandkids Australia from the USA.
		     Accesses     Hosts

 US Commercial (.com)   61.4%     44.1%
 Non-Profit (.org)      18.5%     14.0%
 Network (.net)          9.7%     16.8%
 US Educational (.edu)   8.7%     21.7%
 US Government (.gov)    1.1%      2.1%
 United States (.us)     0.6%      1.4%
Again there is a disparity evident between computers and accesses. In particular, the approximately 39% of computers in the network and education domains accounted for less than 20% of the hits from the USA. It is difficult to put this observation down to network speed, and no satisfactory explanation is currently at hand.

Navigational Indexes
The most common way of reading a conventional book is linearly from front to back. Consequently, it was felt that the linear navigational paradigm should certainly be one of the methods supported in the hypertext version. Following a brief explanation of what the book is about and who the author, Brian Bergin, is, the Grandkids Australia home page has a hypertext link directly to the first section of chapter one. (More precisely, the link specifies parameters to the database search engine which then runs on the server machine, extracts and formats the first section, and returns the formatted page to the client.) Each section in turn has a pointer to the next section (again added on-the-fly by the database software) to facilitate the natural forward navigation.
Each section also has a hypertext pointer to the previous section to facilitate re-reading and backing-up to check specific things which might have caught the reader's attention. In addition, each section has a pointer directly back to the home page to help avoid readers getting the feeling that they are "lost" in a complex hyperspace. Upon return to the home page, readers can access various indexes to get an overview of where they are up to and its relationship to the rest of the book.
Early feedback from a subset of the readers indicated that whereas some were happy progressing section by section, others preferred to read larger chunks such as a whole chapter at a time. Several issues were involved. Firstly, there was the human interface aspect of having to click (and sometimes wait) at the end of each section, rather than only once per chapter. Also, many clicks per chapter generates more network overheads than one per chapter, even though the wait for a whole chapter would in general be longer. The longer wait to download a whole chapter is not a major issue in practice since many web browsers display what they receive progressively, so that the user can begin reading even before the whole chapter has been received. A third issue was that some readers preferred to download a larger chunk for storage locally on their own computer and later detailed reading off-line.
The database method of storing the chapter sections facilitated adding a chapter by chapter navigational method. A hypertext link was added to the front page which pointed to the whole of the first chapter. More precisely, that link contains parameters to the database engine instructing it to retrieve all the sections belonging to chapter 1 and format the whole in accordance with a new page template which is appropriate to a whole chapter. Therefore, the main change needed was the design of a single new page template, rather than having to reformat hundreds of pre-formatted pages.
One key difference between sections and chapters is that there are very many sections in the book, but only 25 chapters. It is feasible to list the numbers 1,2, ..., 25 along the bottom of a page and arrange each number to be a link to the corresponding chapter. This is a more convenient and powerful navigational tool than the next and previous pointers specified by the section template. Here is an example of where the use of more than one template (one template for sections and one for whole chapters) proved invaluable because a different navigational method is suited to each.
In addition to the "jumping-off" points to the first section and the first chapter, as described above, the home page contains four links to four different indexes. There is a traditional index by chapter, and a far more detailed index allowing access to all the individual sections. There is also an hierarchical index which lists the chapters, then the sections of that chapter which is selected, and then the text of that section which is selected. The disadvantage compared to the detailed index is that the hierarchical one requires an extra click to choose an individual section. On the other hand, far less information needs to be transmitted as only the sections of the chosen chapter need be listed. Finally, there is an index of photographs which appear scattered throughout the book. This index is in the form of a "photo album" where each photo is reproduced and acts as a hypertext link to the section of the book where that photo again appears.
In practice, all four indexes were found to be useful, although they were not utilised with equal frequency. The chapter index and photo album proved most popular, accounting for 47% and 34% of the index accesses respectively. The detailed section index was only 13%, and the hierarchical index was a mere 7%.

Interactivity
The nature of the indexes discussed in the previous section and the provision of the additional navigational links as described, give the reader considerable power to browse through the hypertext book, not necessarily in the standard linear order. The following graph shows the frequency with which each chapter was read. The percentages on the vertical axis provide a comparison with the number of readers who read chapter two. This was felt to be a more meaningful figure than had chapter one been used, because there were a large number of readers who downloaded chapter one just to check, presumably, if the book was of interest to them. In contrast, the readers who downloaded chapter two were more likely those who actually read chapter one and decided that the book did hold at least some interest for them.

The graph shows that interest seemed to fall off considerably after the first few chapters, with only about 20% of the chapter two readers getting all the way through the book. On the other hand, about half of the chapter two readers availed themselves of the various navigational elements to jump to the last two chapters. The penultimate chapter is the summary and the final chapter contains some interesting interactive elements which are described below.
Some chapters were significantly more popular than others which tends to attest to the fact that readers made use of the indexes and the ability to easily jump into the book at any place which caught their attention. The photo album may have had some influence with chapters 8, 14, 15 and 18 being relatively popular and each containing a photo which was indexed as described earlier. On the other hand, chapter 10 also had a photo and was not as popular, whereas chapters 7, 9 and 17 had no photos, but were visited relatively frequently. Most likely, readers used the indexes to find topics which were of interest to them.
The following graphs trace the frequency with which various sections were read in the first three chapters. Chapter 1 is naturally quite different since most readers started there and used the first few sections to ascertain whether or not the book was of interest to them. Therefore, the drop-out rate is quite high early in the first chapter. By contrast, chapters two and three were each able to retain at least 60% of those readers who downloaded the second section. These figures accord with intuition and are not surprising.

Apart from the interactive navigation, three other interactive elements were built into the hypertext version of Grandkids Australia. The author presents his vision for Australia and challenges the readers to supply theirs. To this end, an HTML form is built into the relevant section of the book and this allows people to type in their own visions. These visions are added to a database by special software running on the server. The net effect is that the readers' visions become part of the book itself and, because of the dynamic way in which pages are formatted by combining selected data with an appropriate page template, the supplied visions are given a look and feel and navigational links which are consistent with the rest of the book. To all intents and purposes, the reader supplied visions become an integral part of the published book. Indeed, the software also automatically creates an index for the visions.
In a similar vein, one section of the book asks the readers to consider their own responses to a set of questions. The hypermedia version allows readers to register their yes/no responses which are dynamically combined with all previous responses to give a statistical picture of voting on those topics. The results of the dynamic voting form a cohesive portion of the published work.
A final element of interactivity is provided by a link which appears on every section page and every chapter page. This link entices readers to enter into a free-form dialog with the author and with one another. Readers can type in any comments about any section of the book. This on-going dialog also appears as a page in the book for all to read and contribute.
It is interesting that voting on the quiz questions was almost twice as popular as the other two forms of contribution. On the other hand, when it came to viewing the results of others' contributions, the three forms were about on par with one another.

Conclusions
Overall, the implementation of the on-line interactive hypertext book described in this paper was successful. Many people from around the world read part or all of the book and made good use of the indexes and interactive elements.
A key factor of success was the decision to keep the content well separated from the graphical and navigational aspects. This allowed much flexibility for experimenting with different page layouts and navigational paradigms during the design and construction phases. It also facilitated the making of significant changes based on real user feedback from the field. An example was the provision of chapter by chapter browsing in response to requests from some classes of users for larger "chunks" of information.
Another lesson which can be drawn from the experience is to cater for a range of users rather than assuming they all fit into one neat cubby hole. Some users preferred to browse by chapter, others by section. Some made use of one type of index, some of others. Many users liked to provide feedback by clicking on yes/no choices, but others preferred to type in paragraphs by way of their contribution. Providing a variety of facilities for navigation and interaction gave readers the flexibility to use the book in the way which suited them best.

References

[ACM 95]
ACM: CACM Special Issue on Designing Hypermedia Applications, Vol 38, No 8, Aug 1995, http://www.acm.org/siglink/
[Bergin 95]
Brian Bergin: Grandkids Australia, http://www.sofcom.com.au/Grandkids/
[Fraisse 95]
Proc. Second Int. Workshop on Hypermedia Design (Fraisse, S et al, eds) Montpellier, France, June 1995
[Lynch 95]
Patrick J. Lynch: Annotated Bibliography - Graphic Design for the User Interface, Yale Center for Advanced Instructional Media, http://info.med.yale.edu/caim/Biblio_GUI.HTML
[Ockerbloom 95]
John Ockerbloom: The On-line Books Page, Carnegie Mellon University, http://www.cs.cmu.edu/Web/books.html

[Return to Table of Contents]
COPYRIGHT © 1995 by AUUG95 and APWWW95 Charles Sturt University. ALL RIGHTS RESERVED.