Metaphor Development in Public Discourse Using an ARIMA Time Series Analysis Approach

Metaphor Development in Public Discourse Using an ARIMA Time Series Analysis Approach Winnie Huiheng Zeng Dennis Tay Kathleen Ahrens Department of Chinese and Bilingual Studies Department of English and Communication Department of English and Communication The Hong Kong Polytechnic University The Hong Kong Polytechnic University The Hong Kong Polytechnic University 11 Yuki Choi Road, Hung Hom, Kowloon, Hong Kong 11 Yuki Choi Road, Hung Hom, Kowloon, Hong Kong 11 Yuki Choi Road, Hung Hom, Kowloon, Hong Kong winnie-huiheng.zeng@polyu.edu.hk dennis.tay@polyu.edu.hk kathleen.ahrens@polyu.edu.hk Abstract This study introduces a Time Series Analysis approach to metaphor development in a corpus of public discourse as a case study to examine the potential implications for the strategic use of metaphors in discourse over time. The corpus covers 20 years of public speeches by the government leaders in Hong Kong. We conducted an ARIMA time series modeling on the use of the frequently occurring metaphor source domains in the corpus. The ARIMA time series modeling procedures were explicitly presented, and the results were qualitatively discussed with empirical examples. We found that LIVING ORGANISM metaphors demonstrate the clearest usage profile across time, which can be attributable to the progressions of background events in the broad context based on the corpus evidence. In sum, our study emphasizes the Time Series Analysis as a complementary method offering structural insights to the diachronic study of metaphors in discourse. 1 Introduction One primary focus of discourse research is how particular linguistic or discourse features change over time. In diachronic corpus-assisted discourse studies, the most popular way to track changes of variables of interest is to compare their occurrence frequencies between corpora in different time frames for evidence of significant differences (Tay, 2019). Statistical tools for significance tests have usually been applied to measure the degree of associations between variables and time. Recently, a few diachronic studies have emphasized the factor of time series when analyzing discourse over time (Koplenig, 2017a, b; Tay, 2017, 2019, 2021a, b; Burgers & Ahrens, 2020; Zeng, Burger, & Ahrens, 2021). These studies observed internal relationships within temporally ordered observations, known as autocorrelation, referring to the occurrence frequency of one variable at time A having influences on its corresponding occurrence frequency at time B. If we simply calculate the frequencies at times A and B and compare these frequencies using correlation tests (e.g., Pearson correlation analysis) for evidence of significant differences, we would overlook the possible presence of the autocorrelation functions between consecutive observations of the temporal data. The findings of these studies thus questioned the conventional correlation tests, which assume independent observations among temporally ordered data that lead to “incorrect statistical inference where potential effects are meaningless” (Koplenig, 2017a, p. 166). In this study, we apply the Time Series Analysis (TSA) method to discourse research in order to address the limitations of the conventional correlation tests. We conducted a case study of metaphor use in a corpus of public discourse spanning two decades. The corpus includes public speeches delivered by the principal officials in Hong Kong from 1 …

Content truncated to 500 words