(Measuring) Readability: a critical look at sustainability reports through standard formulae and NLP AbstractThis study characterises and problematises the language of corporate sustainability reporting along region, industry, genre and content lines by applying readability formulae and Natural Language Processing (NLP) to a manually assembled 2.75-million word corpus. Readability formulae reveal that, despite its wider readership, sustainability reporting is still a very difficult to read genre, one sometimes more difficult than financial reporting. Although we find no industry impact on readability, region does prove an important variable, with Australian reports significantly more complex than others. These results not only highlight the impact of legistative contexts but also language variety itself as an underexplored variable. Finally, the study lays bare some of the weaknesses of default readability formulae, which are unable to detect American reports' more active language or lower lexical density in UK/EU reports, and demonstrates the merits of NLP in report readability analysis as well as the need for more accessible sustainability reporting.

In this article we tackle the issue of diachronic variation in constructional semantics through an exploration of the (recent) semantic history of the well-established English ditransitive or double object argument structure construction. Starting from the assumption that schematic syntactic patterns are not fundamentally different from lexical items, we will show that -similar to the diachronic semantic development of lexemes -the semantics of argument structure constructions in general and that of double object constructions in particular, is vulnerable to semasiological shifts as well. More specifically, the analysis, which compares data from 18 th -century Late Modern English with presentday English, shows that the double object construction's semantic evolution presents a case of specialization, in which the construction has come to be associated with a significantly narrower range of meanings. It will further be argued that such patterns of semantic change are best captured in a model of argument structure semantics which discriminates between central and less-central or prototypical and non-prototypical uses.

No abstract

This paper examines the online spread of non-standard Dutch in an unexplored communicative setting, i.e. complaint management on corporate Facebook pages. Based on a self-compiled corpus of consumer-company interactions taken from 6 corporate Facebook pages, we investigate to what extent typical features of informal social media communication spill over into more sensitive contexts of complaint management. We do so by mapping the presence and frequency of 27 Flemish old vernacular and 13 new vernacular features (cf. ) in consumer-initiated posts, company replies and consumer-to-consumer interactions. The results show that some, though not all, new vernacular features occur relatively frequently in both consumer and company messages. Consumers predominantly use new vernacular features for expressive compensation, while companies appear to incorporate them in their webcare (especially emoji and English insertions) as functional operationalisations of conversational human voice () and to support their desired brand identity. In contrast, old vernacular features rarely occur in the dataset. This suggests a different status of old and new vernacular features, where the former are deemed less appropriate or more limited in terms of functionality in this communicative setting. Despite the often informal register of the consumer messages, Standard Dutch still seems the preferred or safer option for company-addressed posts while companies, too, cling to Standard Dutch as the benchmark for professional written communication in this online context.

No abstract