| Popis: |
Die sisteemteoretiese benadering tot die literatuurstudie is reeds deeglik gevestig, ook in Afrikaans. Sisteme kan as netwerke gemodelleer word en sedert die vyftigerjare van die vorige eeu is 'n groot aantal netwerkmodelle binne die wiskundige grafiekteorie ontwikkel. Die huidige studie vergelyk die hedendaagse Afrikaanse poësiesisteem en die Afrikaanse literêre sisteem van ander tydperke met twee netwerkmodelle ten einde te bepaal watter netwerkmodel die juisste voorstelling van Afrikaanse literêre netwerke verteenwoordig. Twee van die belangrikste netwerkmodelle in die wiskundige grafiekteorie, naamlik die lukrake netwerkmodel van Erdös en Rényi (1959; 1960 - voortaan die ER-model) en die skaalvrye netwerkmodel van Barabási en Albert (1999 - voortaan die BA-model), word bespreek en met behulp van Pearson se korrelasiekoëffisiënt word die korrelasies tussen die skakelverspreidingspatrone van Afrikaanse literêre netwerke en hierdie twee netwerkmodelle bereken. Daar word aangetoon dat sowel die hedendaagse Afrikaanse poësiesisteem as ander Afrikaanse literêre netwerke akkuraat met die BA-model gemodelleer kan word. Die betekenis van hierdie modellering word bespreek, veral met betrekking tot hoe die groei van die literêre netwerk daar uitsien, dit wil sê, deur Barabási se beginsel van selektiewe skakelvorming op die hedendaagse Afrikaanse poësie van toepassing te maak. The system theoretical approach to literature is well-established (Even-Zohar, 1979; 1990; 1997), also in Afrikaans literary studies (Viljoen, 1984; 1986; Senekal, 1987). Systems can be modelled as networks, and since the fifties of the last century, a large number of network models have been developed in mathematical graph theory. The current study follows Li et al. 's comparison (2013) of network models in comparing the contemporary Afrikaans poetry system and the Afrikaans literary system from other periods with two network models in order to determine which network model constitutes the most suitable representation of Afrikaans literary networks. Two of the most important network models in mathematical graph theory, namely the random graph model of 'Erdas andRényi (1959; 1960 - henceforth the ER-model) and the scale-free network model of Barabási and Albert (1999 - henceforth the BA-model), are discussed. In an ER-model, link formation occurs at random, meaning that there is an equal probability that a link will form between any two nodes. Because of this random process, the link distribution pattern in an ER-model follows a Poisson distribution, which means that the majority of individual data points can be found close to the average and that the average is therefore a good prediction of the number of edges that an individual node will have. In contrast, link formation is biased towards nodes that already have a high number of links in the BA-model, meaning that ties are more likely to form with nodes that already have a high degree. The result of this preferential attachment process is that the link distribution pattern in a BA-model is right-skewed and inhomogeneous, with a small number of nodes having a high number of ties, while the majority of nodes have a small degree that falls below the average. Using Pearson's correlation coefficient, the correlation between the link distribution patterns of Afrikaans literary networks and these network models is calculated. With a correlation of r = 0,942 between the link distribution pattern of the contemporary Afrikaans poetry network and that of the BA-model, and 0,836 < r < 0,986for other Afrikaans literary networks and this model, it is shown that both the contemporary Afrikaans poetry system and other Afrikaans literary networks can be modelled accurately with the BA-model. In contrast, correlations between the link distribution patterns of Afrikaans literary networks and those of the ER-model are shown to be weak, with r = 0,138 between the contemporary Afrikaans poetry network and that of the ER-model, and -0,119 < r < 0,043for other Afrikaans literary networks and the ER-model. The meaning of this modelling is also discussed, especially how the growth of the literary network is informed by Barabási'sprinciple of preferential attachment. It is argued that authors that have been studied frequently in the past are more likely to be studied in the future, and that reviewers who have been asked to review poetry books in the past are more likely to be asked to review a new publication than reviewers who have not been as active in the poetry network in the past. While this illustrates that the literary system is self-organising insofar as decisions about which authors and books to review, as well as which reviewers to use are made from within the system, it is also argued that Barabási's notion of fitness can further illuminate the functioning of the literary system; it is through a combination of extrinsic and intrinsic factors that poets become popular in academic circles. |
