Annotated dataset for the paper: Information structure and the agent-first preference: A corpus-based case study of a free word order language

<p dir="ltr">The paper which is based on this dataset investigates correlations between information structure and constituent order patterns in transitive clauses extracted from recorded texts in Jaminjung-Ngaliwurru, an Australian language (Mirndi family) without a dominant constituent order at clause level. In transitive clauses with two overt arguments, Jaminjung-Ngaliwurru shows a widely reported cross-linguistic preference for orders where the agent precedes the patient. However, it is argued in the paper that the prevalence of agent-first orders is an epiphenomenon of two related preferences: a high probability for topics to be agents, and a preference for a rheme-internal placement of the patient, in contiguity with the verb. The result is that agents, in the case that they are overt at all, tend to be placed in rheme-external position, usually as initial topics.</p><p dir="ltr">The file includes the anonymised metadata for the Jaminjung-Ngaliwurru texts from which the dataset is extracted, and the dataset. The dataset itself consists of all transitive clauses found in the texts, represented in a working orthography with English glosses, an English free translation, an annotation following the GRAID (Grammatical Relations and Animacy in Discourse) conventions supplemented by information structure categories, and further annotations for constituent order and information structure which are explained in the file.</p>