A Study of Shell Nouns in COVID-19 News Reports for Application in Automatic Appraisal of Writer Stance

This project proposes to develop an automatic appraisal system of writer stances and attitudes in news media by way of shell nouns (SNs). As a group of abstract nouns such as fact, problem and suggestion, SNs encapsulates the meaning proposition in their complement clauses. The SN and the complement clause have a characteristic semantic relation of equivalence.The research is motivated by the understanding that SNs represent powerful linguistic expressions of speaker stance and attitude in discourse and, when viewed from the perspective of cognitive construal, afford a great deal of insight into language and thought in general and opinionated linguistic encoding in particular. The current project is additionally motivated by the observation that previous studies have focused on academic discourse and very little is known about their use in the broad coverage of healthcare issues in media discourse as an increasingly important outlet of public sentiment. The study intends to fill this gap and aims to chart an empirically verifiable usage profile that usefully feeds into an automatic stance appraisal system. For this purpose, the project will employ a corpus of media reports covering health issues in general and COVID-19 in particular over the past two years and perform a comparative-contrastive analysis of SNs demonstrating different attitudes across different focal topics such as vaccinations, personal protective equipment, and government recommendations. Methodologically, the present study will adopt a novel approach that treats SNs as grammatical metaphors capable of revealing the speaker’s cognitive construal. In this approach, the SNs will be first classified according to an authoritative semantic scheme of analysis, then mapped to transitivity processes and finally categorized according to a taxonomy similar to that proposed Appraisal Theory.Quantitative analyses such as Latent Dirichlet Allocation and Kullback-Leibler Divergence will be conducted and manual validation performed to identify significant patterns of construal characteristic of attitudinal expressions and contextual topics. Since SNs represent grammatical metaphors underlined by cognitive construal, the project is significant for its application of cognitive principles governing the choice of specific linguistic encodings to evaluate the sentiments of the texts. Results will help to unveil different cognitive patterns of construal in news media leading to better AI systems concerned with sentiment understanding. Findings from the project will also lend insight into linguistic research focused on an improved taxonomy of appraisal.Results arising from the study will be disseminated via academic journals and international conferences.