The fan fiction community represents one which has moved from free tagging metadata to using the folksonomies created from those tags. The lack of machine-readable meta- data may have drawn out the progression but the extended length of time which the community has been using the system more than compensates for any potential slow- ness. We can see how a mature system may behave with the tags feeding the taxonomy which in turn informs the community as to what tags are expected and appropriate vo- cabulary. We also see that where there is no formal structure underlining the vocabulary there is a greater variation in the exact tag string used. While a number of factors affect this, the underlying community agreement on vocabulary and meaning, whether in the community using the term or the one for which it was originally formalised, has a strong effect on the divergence of terms.
Despite the amateur writing online domain being, according to the Shirky (2005) model, an almost perfect example for the type of area that could benefit from free-tagging, on balance it does not seem as if it will fulfil the needs of that community as well as a carefully presented ontology would. We argue that given the current state of tag use, the community could benefit from an extension of the tag system which allows greater definition of meaning. We propose the creation of an ontology which can describe the required concepts with the desired level of precision but which can be linked and integrated with the current system. Having decided to develop an ontology, the other important consideration is how this ontology is presented to the amateur users.
In this thesis I theorise that for a system such as that proposed in this thesis to work, it would need to be capable of hidden so that its complexities are simplified in the interface while at the same time being powerful enough to deal with all the necessary situations that it will need to describe. The idea of allowing community tagging using the restricted definitions allowed by the ontology is an interesting one, and is possible with the system already proposed via the FOP files, which allow personae to either include a machine-readable description of a piece of media or to link to one for either their own or other people’s creations. Whether this would be accepted as a continuation of the current recommendation system or would cause problems with malicious or just differently opinionated taggers is a question that can only be answered through testing. There are also other considerations related to the revelation of sensitive information such
as spoilers which needs to be dealt with in the ontology and beyond that would require careful handling when the descriptions of a piece of media’s contents were correlated. The work detailed in the subsequent chapters represents the first steps towards proving, or disproving, this hypothesis.
Modelling
Describing the Narrative Content
of Media
7.1
Methodology
In the previous section we considered the use of user-generated labels as metadata to describe the content of an item of media. We argued that to fulfil the user requirement for the metadata it was necessary to allow for more precision when desired. We showed that even though there was a shared vocabulary the understood meaning was not necessarily consistent. Having concluded that the community would benefit from the option of more formalised metadata, this section discusses the requirements that such an ontology would need to meet.
Hypothesis: From our investigation of user requirements as detailed in the previous
chapters, it was clear that users were interested in information that went beyond that of basic bibliographic details to metadata describing certain types of events of themes that existed within the media. This information was neutral in that some users desired it as a means of locating media with specific content while others used it to aid their avoidance of the same. We therefore argue that by describing the events that exist within the narrative contained by an item of media we can provide a way for users to improve their options for personal filtering either for search or avoidance of particular types of media.
Methodology: To describe the content of media items we draw on research into narra-
tive theory (Holden, 2003; Jordan, 2004; Chatman, 1978; Bal, 1997) and the information gathered from our case study community to inform our design of an ontology, OntoMe- dia, for the description of narrative events. This ontology represents the cumulation of a collaborative research project undertaken by the author, M. Jewell and M. Tuffield. This chapter lays out our design methodology for the ontology. We go on to provide ex- amples demonstrating how the ontology allows linking between related concepts within
different media objects in a media-independent way and how narrative streams can be constructed from the basic elements within the ontology. In doing so we show how the design of the ontology fulfils the user requirements of our case study community and thus represents a contribution to the domain both in the specific, furthering the needs of the user group under investigation, and in general, creating a way of describing any narrative in a cross-medium manner.