Year: 2025
Type: conference-paper
Abstract: The Givenness Hierarchy (GH) models how speakers signal cognitive statuses of referents in discourse, playing a key role in computational models of situated communication and in applied linguistics.We present an empirical method to evaluate the Givenness Hierarchy using large corpora with coreference annotations.Our findings confirm predicted associations between cognitive statuses and referring expressions across multiple languages, while also highlighting limitations, notably difficulties to approximate the cognitive status UNIQUE and to account for demonstrative noun phrases.Additionally, we demonstrate how coreference data can be used to bootstrap GH annotations, facilitating automatic labelling of cognitive statuses and advancing discourseaware NLP.Finally, we provide conversion scripts to transform Japanese and Korean corpora into CorefUD-compatible formats, supporting broader multilingual research despite current annotation and licensing constraints.Our work bridges theoretical linguistics and practical computational methods, offering a scalable framework to study givenness across diverse languages.
Author: Christian Chiarcos
Institution: University of Augsburg
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4416037546