Contrasting human- and computer-generated english: the case of football match report
- Publisher: Institutional Repository of Vilnius University
This paper aims to compare two kinds of football reports: written by football reporters and automatically generated by a video game. The corpus of real reports consists of match reports from the BBC website, while the corpus of computer-generated language has been compiled from video game Football Manager 2017. The aim of the study is to compare two varieties of football discourse by applying the lexical bundle approach (Biber et al. 2004). More specifically, the analysis involves a comparison of human and computer generated language in terms of over- or/and underuse of lexical bundles on the two corpora. The results show that referential bundles are by far the most frequent in both corpora, followed by stance bundles and an insignificant number of discourse organizers. The study seems to suggest that human reporting has a larger capacity to describe events or object which are outside the scope of the actual match reported and to express stance about the general pattern of play. However, the study reveals some striking similarities between the corpora, which suggests that, in many ways, Football Manager impressively recreates the feel of real life reporting.