Hypertext, 9 Months, and Prototype

Much has occurred since I last updated this blog, so much in fact that I failed to keep this up to date. I intend to work on updating this more but for the mean time allow me to catch up on my work over the last 4 months.

A Thematic Approach to Emerging Narrative Structure was completed and accepted and I took it to the Web Science workshop at Hypertext '08, I also submitted my work on the thematic model to the ACMs SRC as a poster and presentation that was presented at Hypertext. My work received a lot of positive feedback at the workshop and an excellent discussion was had over the nature of web science and how it could become defined. The SRC was also a positive experience which gave me room to discuss my work with many intelligent individuals and take in possible ideas for where my work was going, I also managed to win 2nd place in the SRC itself - which was a pleasant boost of confidence.

Up to the conference I also spent much of my time compiling my 9 month report, essentially a document exposing my achievements and work for the 1st year of my PhD and my plans for the future, this was followed with a viva shortly after hypertext that, while difficult in places, was successful. It highlighted that I need to give careful attention to how my work will be evaluated and how the model itself can be assessed on its own merits rather then the merits of an instance of the model.

Currently I am back at work on the thematic prototype. This small prototype is a basic implementation that uses the thematic model to build themed photo montages (a very basic form of narrative). It was serve as a testing ground to raise any issues that need modifications to be made to the model through some qualitative testing. Currently I am exploring the effects of different metrics for calculating thematic score; the coverage of motifs a natom has, how well matched to a specific motif a natom is, and how focused on a specific motif a notam is. As well as exploring different ways of building instances of the model. So far the act of building the prototype has been extremely enlightening revealing many possible issues with my model and its implementation unrevealed before such as association within themes and whether a relevant theme is a sub theme at all or something the model does not allow for.

I hope further exploration of this prototype to answer several of my questions and to allow us to refine my model.

A Thematic Approach to Emerging Narrative Structure

For the past month I have been working on and developing my thematic approach to narrative. This has resulted in a model of themes within stories which is the center piece of a paper called "A Thematic Approach to Emerging Narrative Structure" that I have submitted to the Web Science Workshop of this years Hypertext conference.

As I discussed before I have been investigating Tomashevskys work on thematics and how it could be approached to the construction of narratives as supposed to deconstruction.

The model details that themes can be constructed from other themes and motifs. Motif's are the atomic thematic elements that cannot be further broken down - they are often generalisations of entities that exist within a narrative and are directly denoted by the features within the contents of a narrative. We use the term 'Narrtaive-atoms' or 'Natoms' to describe the smallest atomic narrative segments that a story is built of such as a photo, or paragraph, it is from the meta data of these Natoms (tags or otherwise) that we build up a list of features for each Natom and through them denote motifs that can connote themes. The diagram below details this model.

Such a model could be used to identify the required features for a narrative to communicate a theme and as such if used to influence the story selection of a discourse in narrative generation to generate richer narratives with direction, flavour, and subtext.

The next step is to work towards evaluating the effect of a thematic appraoch and this system on narratives and narrative generation. To this end I am now working on making a simple montage system that can expose small story selections of narrative atoms and then selections that have been influenced with themes built from this model.

Early Ideas on Narrative and Thematics

I have recently come to the conclusion that it would be inetersting to pursue the idea of modeling Narrative from a thematics perspective. The idea behind this being that if we can communicate ideas through themes that are pevalent through out then this will generate more effective narratives and communicate the implied author more effectively then those that only deal with narrative at a more atomic level.

In his writings on "Thematics" Tomashevsky talks about a motif being the smallest thematic element, that motif's belong to the genre of any one narrative and that through motif's in a story Sub-Themes and Themes may be built. This is, as a computer scientist, a very attractive idea to model as clear objects have been defined within the narrative construct and we can use these own rules in building definitions of elements for a experiment in generating narratives.

These are very early prelimary thoughts but I have begun to think about how a basic schema modeling this might come together. If we make the assumption that we have a selection of narrative chunks, whether they be pictures, text, or video, and that these chunks are tagged based on their content - we could use this meta data to build themes within a generated narrative. We could say that these meta data tags are symbols that signify particular motifs, and in turn certain motif's and other thematic elements could be availble to construct certain themes. So, if we have a Fabula of available narrative chunks and wish the generate a narrative conveying certain themes, we could use this meta data to select narrative chunks with tags symbolising appropriate motifs to contruct the required themes. Each symbol would have a list of tags that relate to it, each motif a list of symbols that could signify it, and each theme or thematic element a list of motifs or other thematic elements that could construct it. These lists could perhaps even be statistically weighted to show, for example, which motif's are more effective at constructing particular themes.

There are many things still to think about with such a system, for example a discourse is not simply fabula selection, it is also presentation of the selection of the fabula. In Tomashevsky's work he refered to these as the story and the plot, the plot being the fully presented story. It would require more work and research to find a way of generating an order and medium of presentation for us to generate a plot and thus a full narrative.

An Adaptive Hypermedia Christmas

As I mentioned in previous posts much of my time at the moment is spent reading around my field, to give me a good grounding in Narrative and ascociated fields. This post is really an update with my progress with regards to that.

I began my review in October by reading up on Narratology and Narrative Analysis and then moved on to more practical papers in Narrative Generation. I looked into several different narrative generation techniques all with different scopes of what could be defined as narrative. This ranged from work done at CWI which sought to use narrative techniques of presentation to create more meaningful search results to entertainment narratives such as Facade and the Automatic storyteller. The later investigated the rules of what creates the most entertaining story to generate an entertaining Narrative either as a story or in the case of facade adapt a a story as it goes based on interactions.

This xmas to continue the spread of reading around my field I attempted to gain an insight into adaptive hypermedia, reading important papers in the field such as HyperCafe and papers on several adaptive hypertext projects that could be useful to my research on generating narrative, particularly AHA! and InterBook.

While I expect my literature review to continue I am now begining to think about applications of what I have learnt, and to tackle a small problem that will work towards the greater problem of narrative generation. Throughout my Narratology study I came accross the idea that an important aspect of analysing narrative lies with the author, whether the author is personified in the text or not every story has an implied author - a voice through which the story is told. Thus to accurately generate a narrative from a computer we must consider or model this author from which the story comes.

An important aspect of this implied author is its human affect on the story. This could be a bias, or how the story is structured or told and this is the area I am currently investigating with a look to devloping a small experiment or application.

Starting work with Narrative

Just a quick post to establish the area I am working on.

To start with, a definition:

nar·ra·tive
noun
1. a story or account of events, experiences, or the like, whether true or fictitious.

I am interested in Narrative (stories) as a representation of Information, and how a Narrative can be generated from other information sources. The applications of such research are as varied as the use of stories today ranging from automatic news readers, analysis of colabritive working enviroments, and adaptive storytelling for games to name but a few.

It is fundamental to my research to explore the existing use of narrative, why we use narrative to communicate information, and how it is built, as well what elements of narrative cause it to be used as a method of communicating information and how can these elements be generated. Much work has been done in the study of narrative and how it is built in narratology and a small ammount of research has been done into systems that generate narrative but the field still has many important questions to be answered.

I have begun my research for the past 2 months with a two pronged approach. On one side I have been researching narratology by reading a collection of writings on narratology called "The Narrative Reader" (M McQuillan, 2000) and a book called "The Classical Plot and the Invention of Western Narrative" (N Lowe, 2000). While simultaneously studying several systems that investigate the generation of narrative such as Artequakt, the Hyperdoc, Topia, and some others. I have also began looking into semantic wiki's.

I intend to use this blog as a means of tracking my research progress and also advertising any ideas or discoverys I have. Hopefully this research can make a significant contribution to the field of Information Representation and Narrative Studies.