1 Introduction & Terminology

1.1 Introduction

 In a hypermedia publication (networked or CD-ROM) one is trying to create something that incorporates a variety of information sources and technologies into a structure (form) that relies on very novel and immature delivery and production tools and techniques. One does not even know what to call the target of one's publication - browser, reader, watcher, navigator, participant or what.
 The production phase involves a team with heterogeneous skills and suits. Along with the usual managerial complexities of project work the team has to overcome and use the unmapped artistic and/or methodological territories of interactivity. Essentially one is building a data/knowledge base which will be wrapped into an interface that is affordable, simple yet challenging and either productive, entertaining, educational or all of them.
 In a truly hypermediaic way one should be able to start off with the 'kick-off memos' and preliminary sketches as a database that evolves and versifies itself during the project to the end title. The size and number of contacts of the team set demands on the flexibility and usability of tools and their interfaces - ideally each member of the team could input to the title and its production via whatever computers/programs available. Use of computers will give the team realistic feel for the media and reader's 'look and feel' while they are working on the title be it in brainstorming, production or publishing phase. In a word, implementing interactive, computer-based collaborative technologies in production will eventually improve the quality of titles and - in the long run - lower production costs by offering access to the skills, potentials and experiences of the team in implementing, learning and using digital media. After all, the end product will be distributed to various computers with heterogeneous audience that hopefully outnumbers the authors.

1.2 Some terminology

 I try to present here a hyperboarding methodology that you could implement on variety of computers and will start off with some definitions. Hypermedia document, hyperdocument for short, needs both structural definition and material description.
 Multimedia is often understood as a composite of multiple media with some user interactivity - mostly multimedia titles are organised linearly, so that there is principally 'one way' from start to finish. Hypertext principle relies on direct access to any references within the text. Web of references forms a hyperbolic structure that is essentially holistic in nature - it is more than the sum of its component parts.

1.3 Hyperdocument contents

 In spite of the risk of sounding obvious I will analyse first the component data types of hyperdocuments. Mainly hyperdocuments may contain following types of media.
 Text is pretty much the same whether it will be spoken or displayed. This is the starting point for the production. Once the text bulk has been delivered it is possible to start refining and replacing it with other media.
 It is often helpful to express everything in writing just to secure the uniformity of design. Every item and its functionality in a hyperdocument can be described in writing - however there are at least three distinct types or layers of text the team needs to produce: narrative, algorithmic and dialogue text. Story lines are essentially the material of your title, the knowledge or artistic experiences are imparted with the help of stories you tell. The algorithmic writing describes the functionalities and workings of your title accurately enough so that they can be programmed and laid out to realise the form of your hypermedia title. Algorithms are often formulated as stories: the dining philosophers algorithm, the travelling salesman's problem, the towers of Hanoi, the two armed bandit problem, etc., the problem or the task the algorithm should solve is represented in a format of anecdote or story. Algorithmic writing is refined writing, one has to line out things in a proper order and describe them precisely enough. The exactness of algorithmic writing is something we are generally not trained to do and we should since we are working with computing devices.
 Dialogue is not just "Cancel" and "OK", instead one has to have some sense of what the user was doing when one gives out the small and precise instructions, warnings, error messages, etc. whatever the immediate user support text might be. Quite a lot of that may eventually be what the program agents are communicating to the end user. Finally overall support material, which one might see as specially oriented combination of all three above. Ideally some technical notes and memos could develop into packaging material and documentation.

Picture 1: Some characterisations and examples of the six media component categories of hyperdocuments.

 Sound is used in hypermedia as an element to create a spatial or emotional atmosphere, or to control and pace the user like rythmically tightening bell sound often used in quizzes when the time is running out. Last but not least sounds are often used as alarming or notifying effects or cues, earcons. An example of an Audio User-Interface design may be found in (Mynatt, Edwards).
 Still images can be hand drawn, scanned, photographed, grabbed from video or visualised from non-visual sources. They may come from a photographic source or they can be graphic, computer or man-made. Important characteristics for the production of images are colour schemes, absolute size, resolution and filetype, these are the properties that need to be decided at the earliest - so that resource requirements (amount of work, computing time and memory requirements) can be estimated.

Picture 2: An example of a computer manipulated photo.

 Moving images means literally anything that moves: multivision, animation, video, morph or model. There are two totally different approaches to movies on a computer, either every image displayed is previously produced or images are computed as they are displayed. The attributes that one needs to decide on are frame rates, frame sizes and amount of colors. Quite often it makes sense to use animation instead of videoclips and cut down memory requirements. Animation is very cost effective and has in many applications even higher information content than video.

Picture 3: A morph, saved as a QuickTime-clip.

 Video just pours information on the viewer who has no control over the rythm nor the flow of information therein. In manual work external pacing is considered bad both in view of health and safety as well as economy and efficiency. Demonstrations introduce interactivity into documents. The browser can pace herself by pushing control buttons and okaying dialogues. Demonstration may be defined as application-driven interaction. Technically demonstration offers a superficial improvement to video and animation, and is heavily oriented to the story and its goals. Very often instructional material is organised as demonstrations because it makes the database cognitively more palatable yet keeping the author's message clear.
 Simulations are the real power of any computer based media. In simulation we have a model that user can manipulate by changing the properties and functionalities of its components to make the model behave differently. Classifications of simulations from pragmatic view and degree of independence offer an insight into variety of related genres: games, intelligent environments, training simulations, virtual realities and artificial life.

1.4 Structural definition

 Hyperdocument realises a hyperbolic structure that can be described as a web of associations. It is said that all pieces of information - all of the human knowledge - constitutes and builds up into a web of knowledge which is highly interconnected forming a hyperspace.

Picture 4: An example of a hyperbolic structure, a hypercube of 4th degree.

 We can think of the whole of human knowledge as a recursive whole. Let us onsider genealogy of sciences: we have mathematics, philosophy, physics, etc. that are split down to families and groups of problem and methodology configurations. The hierarchy of disciplines is often highly interrelated in individual documents. It is not very difficult to imagine that if we take the 'shells of sciences' away and connect the inner nodes of problem and method formulations directly, we end up having the same web - infinite, all-encompassing whole.

Picture 5: Recursive interpretation - a node within a node within a node...

 In my opinion recursive modelling of hyperstructure relates to zooming in and out of the structure according to a formulated principle - e.g changing the level of abstraction. It does not crack the essential nut: high interconnectivity makes it very hard to start writing a hyperboard (storyboard for hypermedia) on visually limited and essentially linear tools. You are entangled in a web of tales and their connections and you do not really know how to lay them down 'as you thought' and in collaboration with your team.

Picture 6: a diagrammatic representation of multiple choises in Montage en Dix Parts

 Interactive piece Montage en Dix Parts: I present you a stack of ten photographs. You shuffle them and then read them: the sequence of ten makes up one story. You can reshuffle and read for as long as you like. With only ten cards, one can come up with plethora of over 3.6 million sequences. The artistic challenge lies in finding the proper pictures to carry enough meaning and interest.
 Ted Nelson started the discussion about framing or chunking of information, how a human being according to her interests and abilities frames or chunks out parts of hyperdocument - as well as her own sensory surroundings. (Nelson) Whenever a user is navigating through this web, she always has some sort of frame of concentration, which she understands: I am here, this is a whole, those are parts of that and so on. In real world we perseive and distinguish objects and their relations - this is also true of abstractions like thoughts and hypermedia. These relations fall into two categories: is-a-part-of and is-a-kind-of. Hierarchies are inherent in former category and often very well understood by any body. But what about collections and sequencies of is-a-kind-of entities? I claim that the natural way for us is to organise these into causalities, narratives and spacial or functional structures. Hyperdocument can be defined as a collection of is-a-kind-of objects that represent collections of is-a-part-of objects that are framed out for the reader. Because links representing both relations (narration and dialogue) lie in the same visual whole hidden behind the hot spots, it is very easy to confuse them. It is the task of author to provide the reader with adequate cues to differentiate them.

1.5 Hierarchies, trees and stars

 Ben Shneiderman (Shneiderman) promotes hierarchical linking of information in hypermedia, personally I am very sceptic about using hierarchies in descring relations between is-a-kind-of entities. Hierarchies often lead to a fallatic structural interpretation of a whole that is not truly hierarchic.
 Game trees are very useful in explaining this. Take for instance the good old tic-tac-toe. Two players place X:s and O:s on a board of three by three squares - winner is the one who gets three marks in a row: first plays cross. The game tree is contructed as a tree representing all the possible turns of game from empty board on to victory or loss. The cross has been played in the centre. Of course, there are other alternatives, but I have chosen this path. The next move is the zero, and I have drawn part of the tree, so only portion of the alternatives are here. Ideally I should draw a complete tree with weights for every choice so that the playing algorithm identifies the node representing the game situation and chooses its move according to the weights - heavier the move, surer the victory.

Picture 7: Partially drawn game tree for TIC-TAC-TOE. The colored dots represent similarities in branches.

 A complete game tree is redundant, ie. some branches are equal, some branches are projections of others etc. When redundant branches are unified tree- structure is transforms into a graph. Similar structural difficulties rise if the player is allowed to undo moves.The problem is that when you take one move back it is still acceptable, but if you are allowed to retrace a couple of more moves, then the nice hierarchical organisation falls apart and becomes a graph.
 If you are building a hypermedia document or a computer game you may want to start by constructing a game tree. However it is important to realise that you are working with a hyperbolic structure - a hypergraph that you want to constrain (aesthetically) so that it presents navigational sequences that unfold over time.
 Finally we have centered and starlike structures: consider a map of Helsinki with interactive pop-up explanations of sights and buildings and a visual dictionary. One could safely call that a star structure with a nucleus for an index to the whole document. With star structures one ends up navigating in graphs only crossing the nucleus or index more often than other parts of the document. One cannot get rid of hyperbolic graph structures no matter what.

1.6 Universal access is easy

 All hypermedia environments support linking of screen items to named document parts. Algoritmically universal assosiativity is easy to implement by naming the parts logically, descriptively and unambiguously. Any programmable screen object with the same name can perform a procedure such as "go to the part with the same name as me" - if we know a part's name we can create access via representation to it.

Picture 8: Association through reference by name is the basis of all hypermedia.

 I want to underline that it is not difficult to implement universal linking, the real trick of a hyperboarder is to name all components in is-a-part-of and is-a-kind-of categories so that they relate meaningfully to the dialogue and narrative.

Contents Next