In Context Verb Disambiguation: A Korean Case Study With Reference to Czech

Over the past 20 years, strategies languages use to manage the distribution of information within and across words, sentences, and even broader domains have become an active area of linguistic research. Studies range from distribution on sounds and speech rate properties (Pellegrino et al., 2011) and morphological system organization Ackerman and Malouf (2013); Filipović Ðurđević and Milin (2018) to sentence properties (Futrell et al., 2015; Levy, 2008; Levy and Jaeger, 2007; Wasow et al., 2011) and even studies of broader speaker interaction (Depperman, 2015). One particular recent focus has been the relation between word class properties and their relation to language learning and sentence processing, specifically from a perspective of uncertainty reduction. Dye et al. (2018) offer an analysis of noun classification in languages with (German) and without (English) gender and identify two major strategies for reducing uncertainty: a deterministic system and a probabilistic system. Existing studies in this area have primarily focused on noun classification. Verbs and their interaction with argument structures, which have been extensively discussed in the theoretical and applied literature, including in work incorporating suggestions from Levin (1993, 2015), have received less attention within quantitative literature focusing more directly on cognitive dimensions, and specifically on how verbs and argument structures interact with verb classification. This thesis uses quantitative metrics based on distributional semantics to explore how argument structures and verb meanings interact and define argument-structure-based verb classes. The focus of this thesis lies primarily on Korean, a strict verb-final language, but results for Korean are compared to Czech. In terms of argument structure marking and argument expression, Czech and Korean share a range of features, but contrary to Korean, Czech has flexible verb placement. It is used as a counterpoint to further assess whether the results found for Korean are specific to a language with strictly clause-final verb placement or whether some of them may be more general. The results of this thesis strongly suggest that (at least in two typologically different languages such as Korean and Czech) verbs generally follow a probabilistic classification strategy (in the sense of Dye et al. (2018)). Verbs in both Korean and Czech appear to be classified by their argument structure in a taxonomic way and do not replicate partial idiosyncratic patterns found for nouns in high frequency contexts (Dye et al., 2017; Wang and Walther, 2023a). By highlighting the complex interactions between frequency, semantic similarity, and argument structure, this thesis provides insights into the role of argument structure alternations (including argument omission) in verb disambiguation and classification.

Author

• Cutter, Adam A.

• Quantitative Linguistics
• Korean Linguistics
• Syntax
• Argument Structure
• Language and Cognition
• Information Management

Identifier

hdl:1920/14137

OAI identifier oai:identifier

oai:MARS:1920/14137