An LLM-Assisted Study of Attitude Construction in News Discourse: Evidence from the 2025 Sino-U.S. Trade Disputes
DOI: https://doi.org/10.62517/jnme.202610109
Author(s)
Tianran Wang
Affiliation(s)
Jiangsu Food and Pharmaceutical Science College, Huai’an, Jiangsu, China
Abstract
Drawing on the Attitude subsystem of Appraisal Theory, this study compares how evaluative meaning is constructed in early English-language reporting on the 2025 Sino-US trade disputes by two influential outlets: China Daily (CD) and The New York Times (NYT). Using Factiva, a comparable corpus of 82 reports (42 CD; 40 NYT) was compiled to represent the initial stage of the dispute. Methodologically, the study operationalizes Affect, Judgement, and Appreciation as an annotation scheme and employs a large language model (LLM) as an automated coder to extract attitude resources, label polarity (positive/negative), and identify the speaking actor in quoted (citizen/official/expert). Prompt design was refined through a pilot calibration against manual coding, yielding high agreement (precision 0.91; recall 0.88). On the full corpus, the LLM identified 473 attitude resources (CD 215; NYT 258). Results show convergence in attitude type distribution: both outlets rely most heavily on Judgement, indicating that trade actions are primarily moralized through responsibility and fairness claims. However, sourcing and polarity patterns diverge: CD attributes evaluation mainly to officials, whereas NYT foregrounds citizen voices; and NYT exhibits a stronger negative tilt, while CD displays a comparatively more constructive orientation. The paper argues that institutional role, audience design, and genre constraints jointly shape how trade disputes are personalized and framed. It also outlines a replicable LLM-assisted workflow and discusses reliability and error auditing for appraisal research.
Keywords
Large Language Model; Appraisal Theory; Attitude Resources; China Daily; The New York Times; Trade Disputes
References
[1]Read, J. and J. Carroll, Annotating expressions of Appraisal in English. Language Resources and Evaluation, 2012. 46(3): p. 421–447.
[2]Pang, T.T.-Y., Leveraging large language models to supplement corpus-based inductive learning of Chinese as a second language. Applied Corpus Linguistics, 2026. 6(1).
[3]Martin, J.R. and P.R.R. White, The language of evaluation: Appraisal in English. 2005, London: Palgrave Macmillan.
[4]Ferrari, M., F. Kurcz, and M. Pagliari, Do words hurt more than actions? The impact of trade tensions on financial markets. Journal of Applied Econometrics, 2022. 37(6): p. 1138–1159.
[5]Brown, S., Free trade, yes; ideology, not so much: The UK’s shifting China policy 2010-16. British Journal of Chinese Studies, 2018. 8(1): p. 92–126.
[6]White, P.R.R., Praising and blaming, applauding, and disparaging – solidarity, audience positioning, and the linguistics of evaluative disposition, in Handbook of interpersonal communication, G. Antos and E. Ventola, Editors. 2008, De Gruyter Mounton: Berlin. p. 567–594.
[7]Taboada, M. and M. Carretero, Contrastive analyses of evaluation in text: Key issues in the design of an annotation system for attitude applicable to consumer reviews in English and Spanish. Linguistics and the Human Sciences, 2012. 6(1-3).
[8]Thompson, G., Affect and emotion, target-value mismatches, and Russian dolls: refining the Appraisal model, in Evaluation in context, G. Thompson and L. Alba-Juez, Editors. 2014, John Benjamins: Amsterdam. p. 47–66.
[9]Gao, Q. and D.W. Feng, Deploying large language models for discourse studies: An exploration of automated analysis of media attitudes. PLoS One, 2025. 20(1): p. e0313932.
[10]Križan, A. and A. Barbič, Appraisal Analysis and AI Chatbots: Do We Even Need Humans? ELOPE: English Language Overseas Perspectives and Enquiries, 2025. 22(1): p. 35–52.
[11]Tian, L., et al., Task and Sentiment Adaptation for Appraisal Tagging, in Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics. 2023. p. 1960–1970.
[12]Pounds, G., Attitude and subjectivity in Italian and British hard-news reporting: The construction of a culture-specific reporter voice. Discourse Studies, 2010. 12(1): p. 106–137.
[13]Bell, A., The language of news media. 1991, Oxford: Blackwell.
[14]Lee, G., Verb objectivity and source qualification: Comparison of quotation attributions in offline and online newspapers. Journalism, 2016. 18(7): p. 890–906.
[15]Sundar, S.S., Effect of source attribution on perception of online news stories. Journalism & Mass Communication Quarterly, 2016. 75(1): p. 55–68.
[16]Dür, A. and B. Schlipphak, Elite cueing and attitudes towards trade agreements: the case of TTIP. European Political Science Review, 2020. 13(1): p. 41–57.
[17]Kleemans, M., G. Schaap, and L. Hermans, Citizen sources in the news: Above and beyond the vox pop? Journalism, 2016. 18(4): p. 464–481.
[18]de Hoog, N. and P. Verboon, Is the news making us unhappy? The influence of daily news exposure on emotional states. Br J Psychol, 2020. 111(2): p. 157–173.
