What makes poetic language different? An experiment in genre recognition using word co-occurrence networks in Afrikaans

Abstract

Despite being one of the oldest literary genres, poetry is notoriously difficult to define. Nevertheless, various literary scholars draw attention to poetry’s foregrounding of language itself, as opposed to using language to convey meaning. In the current study, we investigate whether there are structural differences between Afrikaans poetry and prose when texts are analysed as word co-occurrence networks from the perspective of network science. Specifically, we investigate average path lengths (L), clustering (C) and the small-world index (S). We find that poetry texts generally have higher average path lengths and lower clustering than prose texts, which results in lower scores on the small-world index on average. We also calculated these scores for 100-word text segments to eliminate the possibility that text lengths may affect our results, with the same results, albeit with a smaller margin. This means that poetry is a less densely connected genre than prose. Suggestions are also made for further research.