Projects // Contextual Browsing Language (CBL)

CBL and strategies of automatic storytelling

Markup-language, which is based on XML, provides a basis for a telling system that creates visual illustrations of a complex story according to the recipients’ interests. It also forms an evolving story out of an unstructured collection of text parts. The Automatist Storytelling System, which is here, introduces a possibility to construct such a system. Thereby, the term “automatism” can indeed be used in a surrealistic sense, which means as an unconscious or intuitive movement of association that is deployed by keywords.

A communication based on web-service allows an exchange of information. Yet, such an exchange of information will fail if for example XML-tags have identical characteristics, but different names, like for instance <zip>50500</zip> and <zipcode>50500</zipcode>. To avoid difficulties like that, the Semantic Web introduces a meta level, where the meaning is described. There are different ways to do so. Two easy ways to avoid difficulties of meanings are taxonomies as well as classifications. Thereby, every information unit gets classified and thus builds a node which is either graduated to a higher or lower level. Concerning their structures, taxonomies are limited to illustrate a usual hierarchy.

Considering the illustration of logical theories, these approaches are not sufficient. These theories consist of axioms, which lead to certain conclusions according to assumptions and rules that are said to be veritable. Relations between such information units are complex by nature and can be illustrated by ontology in the Semantic Web. Contrary to the philosophic field of ontology, the technical field is far more concrete. It is about vocabulary and the corresponding relations, rules and concepts for a more detailed description of a defined field of knowledge. Thus a machine “learns” to understand what was “meant” by another machine.

Strictly seen, the Semantic Web is not a specification, but rather a philosophy. Due to the W3 Consortium, there is an agreement which includes techniques that are constituted by the Semantic Web. Meanwhile, a lot of specifications exist to reach the target. Basic techniques such as URI, XML and RDF have been developed independently from the efforts of the Semantic Web. Often they are even older. Nevertheless, all theories of the Semantic Web do not describe how to apply to the new knowledge. CBL as well as strategies of “mathematical” storytelling imply a possibility to transform a collection of documents into personal experiences.

The interactive narrative has the potential to tell more complex and personally meaningful stories than those delivered to a mass audience.

Michael Murtaugh

The Markup-Language CBL is based on considerations taken from a paper on “Automatic Storytelling” written by Michael Murtaugh from the Massachusetts Institute of Technology (M.I.T.). He describes a system of telling which forms the visual illustration of a complex story according to interests of recipients. Furthermore, the system changes an unstructured collection of text parts into an Evolving Story. The interface takes a special position considering the editing of the story. The Apple Hot Sauce-Browser was an early approach to visualize data connections.

Fig. 1: The Apple Hot Sauce-Browser

Fig. 2: Plumb Design Visual Thesaurus

In his paper, Michael Murtaugh describes a system of dynamic connections, which consists of keywords and which evolves into a unique text organism regarding the readers’ interests and knowledge.

CBL describes structural connections between data and in addition defines tags to manage the visualization of data. The descriptions of data connections provide a contextual editing of data. Furthermore, a view on the elements of a story, a website or a complex collection of knowledge, which is accommodated to the user’s habits.

Items depict morphemes (the smallest sensible units) in a CBL-Document – they correspond to a text, picture or a URL. Sequences are compact and successive series of items. The Narrative Engine, which calculates the succession of content elements according to the interaction of users, does not have an impact on the succession of elements here. The definition of a sequence ensures the integrity of succession. CBL works according to the “Pickup-Principle”. A CBL-Browser or a Rendering Engine picks only the information from the Item-Elements that are needed to visualize the data, for example labels or picture information.

CBL provides two rating mechanisms to make a presentation of information possible, which depend on the user conduct. Firstly, via the rating-attribute of an item: a simple numerical value, which makes certain items seem more valuable than others. It depends on the CBL-Browser conversion, if the item looks bigger, closer or brighter. All related elements are connected to the active element via the jump-attribute. Secondly, via the narration/topic element: items announce their vicinity to a certain keyword via a rating-value, so to an abstract topic like “Politics”. Due to this detour – the definition of interests replaces concrete links – a dramaturgical reconditioning of data in regard to Murtaugh will be possible. The Narrative Engine gets a feedback by the activity of users. Therefore, a continuous offering of more elements, which are suitable for the user’s interests, can be used. Thus, it is possible for the structure of a website to be as Michael Murtaugh’s Evolving Story says: texts, links and pictures are shown according to the users’ interests. Thirdly: The possibility to get a presentation according to the user’s conduct is off CBL. These strategies only need bidirectional links of content elements as well as describing elements within a CBL-document.

The Mature User

Interactive telling tests the break with static, always unchangeably presented story in favor of the openness it actually has. This openness is converted to a constant pattern by the involvement of recipients. The advantage of interactive telling is based on the user, who gets out of his immature “existence of point-and-click” and thus a one-dimensional thinking is replaced by alternative thinking. Murtaugh’s argumentation reminds of the heavy praised creativity of the so called Wreaders. Current views are still far away from this kind of reception.

During the last one hundred years the smallest pieces of film and audio were collected to produce a single program, a radio-show or a film for us. Therefore, an author arranged content elements for us, so that the use, which he aspired, was reached best. Decisions concerning cutting were made according to criteria like continuity and/or suspense. Not least, a whole branch of industry that only deals with the issue of how to index videos has been developed out of the immensely increasing amount of digital content elements (just think of the photo-collection on your PC!). The efforts go from a completely automated indexation to an annotation of data by a human user. When the different elements of media are recorded, they can be found by more or less complex queries. This “retrieval by query”-approach may satisfy a couple of users, but it will not succeed if the user does not know what he wants to search exactly. In addition, we all have different views on our everyday claims.

Evolving Documentary

The creation of a database, which can be extended, needs a certain frame of organization. What exactly will be collected and under which conditions will a content element be added to the database? Which consequences does it have? How is the database organized? The conflict during the reconstruction of the Munich football-stadium is an example that explains the model of Evolving Documentary. For sequential media like TV and radio, the author’s work looks as follows: The first step is to collect material such as comments by the management of the operating company and members of the city council. Afterwards, pictures and audio-material are accumulated. The author creates a report out of the content elements deriving from the accumulation. The most important comments and pictures are brought into an order and accommodated to the length of a contribution. Figure 3 illustrates this process and shows how the author filters the basis of material for the contemplator.

Fig. 3: The current model

The procedure for an Evolving Documentary is different. Each content element which is added to the data basis changes the author’s and contemplator’s knowledge equally and clears the way to the “Big Picture”. The sequential contribution which has just been described is only a small part of a much longer story. It is a necessity to create such an Evolving Story without any difficulties in adding new material of the data basis. In addition, the browser for this documentation needs to be able to supply material, without the contemplator knowing what can be found. A scalability of the data basis as well as room for description are basic technical conditions for such a scenario. Figure 4 shows this process.

Fig. 4: Evolving Documentary

The “catch wording” of single content elements is a critical aspect of the Content-Model. Knowledge Management Systems are often based on typical journalists’ questions (Who? What? When? Where?), but “Why?” is the more interesting interrogative in our context. Unfortunately I cannot offer feasible solutions for its evaluation.

In the model of Evolving Documentary, the contemplator gets a new role and furthermore the author’s work is going to change. The author still maps different content elements, but in addition he prepares the according room of description, which provides the possibility to show coherence between the elements. The mapping of keywords to content elements is a complicated issue: If there are a lot of authors and topics, the keywords get weaker, so that there is a reduction of the storytelling quality. Figure 5 shows the author’s work for an Evolving Documentary.

Fig. 4: The author’s work

Commenting Content Elements

The first step towards the development of a database of content elements is to create a set of descriptors, thus describing elements. The descriptors were separated into four categories, which proved to be positive for the work on CBL. The four main categories are: person, time, place and theme, whereas the theme is a combination of “What?” and “Why?”. The aim of this process is to get a complete set of abstract ideas and elements which are important for the story to tell. If there are too few describing elements, the CBL-browser cannot differentiate elements that belong to a certain theme. If there are too many describing elements bunching together, the browser cannot establish a connection between the single elements. Defining descriptors that fit becomes a basic task for the production of a working presentation system.

The second step towards the development of a database of content elements is to attach the present content elements to the descriptors that have been developed before. A bidirectional linking of elements and descriptors is the easiest way. The author associates a content element with all descriptors that are adequate for the element and attaches the descriptors to the adequate content elements. The links are not assessed for the easiest scenario and therefore the equivalent for an easy keyword-system. To consider the example of the Munich football stadium again, the descriptor “politics” would be linked to interviews with the local politicians as well as with the EU-commissioners who are in charge. CBL provides appropriate tags for such links that can be read and converted by a CBL-Browser. The weighting of descriptive elements is a result of an ongoing process, which will be specified in the following chapter.

Methods of Presentation

First of all, it is necessary to separate decisions for a sequential succession of the content element from the element itself to achieve an automated dramaturgical effect of the presentation. Few simple methods of storytelling help us for later presentations. All these methods rely on the basis of content elements and descriptors, given by the author. Additionally, these methods have to be capable of being influenced by the user. The purpose is to achieve a story created by a user through a combination of both aspects. The story contains a narrative coherence and a significant structure despite its self-determination.


Continuity is a basic principle of cinematic telling. The importance of successively produced content elements is achieved by the mode, how single content issues relate to previous content issues. The ability to create a meaningful view of an entire story plays an important role concerning a coherent plot of a story. The so called Description Feedback of the Contextual Browsing Model ensures continuity.

The idea of Story Context takes up an essential part within the Context Browsing. Such a Story Context is defined by several descriptors, whose importance is represented by the numerical value in the current context. Looking at the example of the Munich Stadium: If the contemplator is more interested in political matters, the interview with the local politician has a value of 100, the saying of the EU-commissioner has a value of 80 and so on. The CBL-Browser calculates a total out of these values and defines which element is shown next. The procedure is illustrated in figure 6.

Fig. 6: Description Feedback

Afterwards, the element is irrelevant for the current telling context, because it has been shown already. In case of several winners the decision is left to chance. If a certain content element has been shown, the relevance of the connected descriptors is increased by the CBL-Browser, while the relevance of all other descriptors is reduced. Thus, the story is directed according to the topic the author’s requests. So the choice of one content element sets further choices. This is called Description Feedback because the choice of one content element affects all descriptors and therefore affects the telling context directly. Content elements work as a bridging between the different telling contexts. During the process, a story is deployed according to the user’s interests.

From general to specific

A further fundamental principle of storytelling is the development “from general to specific”. To be able to conceive the whole context, general and introducing information is given to the traditional media, before specific aspects of a theme are considered. The introducing information builds the frame story, in which specific statements can be understandable. This mechanism is relatively easy to realize, compared to the difficult ensuring of continuity via description feedback. It is a precondition that the number of descriptors, which are linked to the content elements, can tell how specific an element is.

Fig. 7: From general to specific

To examine more general and more specific aspects, we only need to divide the quotient of the total of points of the descriptors which have been associated. We eliminate content elements that have many descriptors, and favor content elements, which are described less and therefore probably have a more general value. After having played the general content, the element gets eliminated from the content pool. The content elements which have not been taken into consideration before can now be re-evaluated due to their importance. Considering the example of the football-stadium, all content elements would be linked to a contribution called “Munich”, which could be shown at the beginning.

In this case, a contribution can be selected especially according to the author’s space of description. An incomplete sentence of descriptors undermines this principle of mathematical storytelling, because links that are not applied by the author are not taken into consideration for an evaluation.

Narrative space

Stories are often evaluated according to their narrative pace. Therewith, the presentation pace or a tedious voice are not meant, but in fact the development of a story. The narrative pace plays a rather minor role considering the world outside the blockbuster cinema, especially in documentaries. For our purpose we declare the narrative pace to be a term, which gives us information, to what extend certain aspects of a story are developed, or according to our telling model: to what extend the telling context ranges within the space of description.

Although the term space of description has often been mentioned in this contribution, it has not been explained sufficiently yet. We can firstly talk about dimensions when we think of the terms within the categories of our descriptors. Our model does not contain a link between the single descriptors neither does it have associations between the single content elements. But how does a CBL-Browser notice the topical vicinity of football-stadium and Allianz-Insurance-Company, which is the eponym of the stadium, without being explicitly linked? Since continuity can be achieved by analyzing the linking of content and descriptors, the inverse view of descriptors and their linking of content elements can be helpful to notice the topical vicinity of two aspects. The existence of one content element, as for example a computer animated version of the new stadium with a logo of the Allianz-Insurance-Company, could employ the necessary connection by an accurate indexing of the data basis. Elements that are not applied in a system cannot be illustrated by the browser.

Fig.8: Finding topical connected aspects by Content Feedback

By analogy with the Description Feedback the process of Content Feedback will derive, which admits a certain motion within the space of description. Descriptors, which are topically close to the ongoing telling context, can be adducted to increase the narrative pace. To avoid a return to already-shown telling fragments, a simple mechanism that works by analogy with the “already-shown-mechanism” of content elements is sufficient. Furthermore, this simple mechanism needs to imply only those descriptors, which were not considered for the computation of the narrative pace. One possibility is a simple decision of the user, as for instance a button called “Push the story forward”. There is a further possibility for contemplators, who want to deal with the stored data first. Firstly, the Engine shortly points out all aspects at a high narrative pace and afterwards the user is given the possibility to go deeper into the story.

If and to what extend the combination of the techniques designated before is useful, to create a narrative coherence and a telling structure out of an unstructured collection of content elements, will be shown by the implementation of a CBL-Browser. Areas, attached by an author, which adapt more techniques than others are possible, too, e.g. the described mechanism for narrative pace could be applied on a subset of descriptors only. Or the contemplator him-/herself defines how the engine acts in certain contexts. The possibility for a manual weighting of content elements regarding the field of “place” combined with a faster narrative pace, e.g. for the subtopic “persons”, would result in a very personal story, where local celebrities talk about their experiences with football players in Munich. It will take a long time until the effects of different combinations of the techniques, which have been introduced, are fathomed completely.