Dataset Discourse Management Constructions in Wikipedia Talk Pages

They involve the same fixed pattern, combining the same categorical slots. However, they diverge in collocational preferences, which reflect functional differences. The raw dataset consists of a table, with each row containing one corpus occurrence as well as the lexical filler items of the categorical slots that recurred in both patterns.

Those filler items comprise a) the connector slot with the connectors wo or wenn, b) the subject slot that in the vast majority of the cases contains a personal pronoun, c) the adverb slot, d) the preposition slot, e) lemmatas occurring in the noun slot that is embedded in a prepositional phrase, f) punctuation marks. These variables are the basis for the collocation measures presented in the paper.

Methods

The data come from a sub-corpus of the German Reference Corpus (Deutsches Referenzkorpus, DeReKo) hosted by the Leibniz Institute for the German Language (Mannheim), which consists of all editor discussions associated with specific Wikipedia articles between 2002 and 2017 (ids-mannheim.de/digspra/kl/projekte/korpora/archiv/wp/). In this dataset all occurrances of discourse management markers such wo wir gerade dabei sind or wenn wir schon dabei sind were extracted and subsequently analyzed.