Congist’24 Social Sciences and Artificial Intelligence: Theory and Practice

The Image of Chinese International Students on Chinese Social Media Zhihu: A Comparison Between Corpus-Assisted Critical Discourse Analysis Approach and AI-Assisted Corpus Approach

18 Ara 2024 15:40

30dk

A7 (Istanbul University Faculty of Letters)

Oral Presentation Case Studies and Applications Session 2.2 (Day 1)

Konuşmacılar

Yanni Sun (King's College London)

Açıklama

COVID-19 and the surging nationalism and populism sentiments made Chinese international students (CIS) targets of online vigilantism on Chinese social media and they face alienation in the homeland (Gao, 2022) apart from discrimination overseas (Russell, 2020). This research examines the image of CIS on Chinese social media through corpus-assisted critical discourse approach. The Discourse-Historical Approach (Wodak, 2015) commonly used for media presentation studies is adopted to theorise and categorise the findings.

328 posts of 280995 Chinese characters on Zhihu (similar with Reddit/Quora) were collected. Major referential expressions of CIS were identified and classified by browsing the general word and keyword lists and examining their concordances. Predication analysis based on concordances of the most frequent referential expressions of CIS in the corpus 留学生(们)international student(s). Chi-square tests were conducted to identify significant differences between CIS and other Zhihu users in referential and predication expressions.

It is found CIS were alienated and stigmatised as the problematic “other” through frames of trouble or degenerate, meritocracy, nationalism, populism, collectivism, and misogyny in the corpus though some comments try to challenge those frames and depict CIS as well-behaved people, victims, the socio-culturally marginalized, patriots, ordinary people without privileges or high socioeconomic status, talents, individuals with rights, and cosmopolitans. Comparison between CIS and other Zhihu users reveals both groups use stigmatising discourses against CIS. Apparently, tensions not only exist between CIS and non-CIS but also within CIS. The major difference is that CIS are more likely to object to the “trouble or degenerate” and “meritocracy” frames, present CIS as “socio-culturally marginalized or isolated”, recount reverse culture shocks for CIS, and depict CIS as cosmopolitans while non-CIS group is more likely to oppose the “victim” frame, stigmatize CIS as trouble or degenerates, position them in a meritocratic hierarchy, and perceive them from a collectivism (especially pro-collectivism) stance.

To evaluate generative AI’s performance in supporting automated qualitative analysis within corpus-assisted discourse analysis, similar referential and predication analysis based on ChatGPT was also conducted. Results were compared against findings above. It is found that it performs reasonably well in identifying and categorising referential expressions of CIS but the predication analysis (concordance analysis of the most frequent referential expressions of CIS) is not very accurate.

Institution / Affiliation / Kurum

King's College London

Sunum Materyalleri

The image of Chinese international students on Chinese social media Zhihu.pdf