Measuring public opinion on the import of US pork in Taiwan

28795

Authors

nan

Abstract

As the import of pork with ractopamine from the United States (US) has stimulated a heated discussion on social media, the purpose of this exploratory study is to assess public opinion on the import of US pork by examining the interplay between the two sub-issues, namely food security (FS) and political economy (PE), at the word, sentence, and time series levels. First, we collect comments related to the US pork issue on PTT, the largest online discussion forum in Taiwan, throughout the year 2021. Second, at the word level, we apply Latent Dirichlet Allocation (LDA) to extract the main topics. Third, at the sentence level, we sample a subset of data for manual annotation. Taking the annotated subset as the gold standard, we use the pretrained BERT-base-Chinese model for the downstream topic classification task to predict the whole dataset. Fourth, at the time series level, we employ Detrended Fluctuation Analysis (DFA) to evaluate the predicted number of comments for FS and PE sub-issues. Two main results are obtained: (1) the theme of political economy interweaves with the concept of food safety at both word and sentence levels; (2) DFA shows the absence of a more salient perspective between the FS and PE sub-issues at the time series level. 1

Keywords

nan

Full Content

Measuring public opinion on the import of US pork in Taiwan Yu-Lun Huang Yu-Yun Chang National Chengchi University, Taiwan 110753134@nccu.edu.tw National Chengchi University, Taiwan yuyun@nccu.edu.tw Jyi-Shane Liu Chang-Yi Lin National Chengchi University, Taiwan liujs@nccu.edu.tw National Taiwan Normal University, Taiwan 61015003E@gapps.ntnu.edu.tw Abstract As the import of pork with ractopamine from the United States (US) has stimulated a heated discussion on social media, the purpose of this exploratory study is to assess public opinion on the import of US pork by examining the interplay between the two sub-issues, namely food security (FS) and political economy (PE), at the word, sentence, and time series levels. First, we collect comments related to the US pork issue on PTT, the largest online discussion forum in Taiwan, throughout the year 2021. Second, at the word level, we apply Latent Dirichlet Allocation (LDA) to extract the main topics. Third, at the sentence level, we sample a subset of data for manual annotation. Taking the annotated subset as the gold standard, we use the pretrained BERT-base-Chinese model for the downstream topic classification task to predict the whole dataset. Fourth, at the time series level, we employ Detrended Fluctuation Analysis (DFA) to evaluate the predicted number of comments for FS and PE sub-issues. Two main results are obtained: (1) the theme of political economy interweaves with the concept of food safety at both word and sentence levels; (2) DFA shows the absence of a more salient perspective between the FS and PE sub-issues at the time series level. 1 Introduction In Taiwan, social issues are typically divided into different discursive dimensions according to political ideology. Take the US pork issue as an example, the Democratic Progressive Party’s (DPP) decision to lift the restriction on the import of pork with ractopamine was made to strengthen Taiwan’s economic ties with the US, while the Kuomintang Party’s (KMT) opposition to this policy heightens the potential risk to food safety and facilitates the holding of a national referendum banning the US pork importation. As the referendum is intended to represent the views of citizens for involvement in the policy-making process, the question then arises as to which perspective the political parties attempt to shape public opinion from. As social media mining has become a powerful tool to investigate the formation and change of public opinion (Anstead & O’Loughlin, 2015), this study evaluates the evolution of public views via the Professional Technology Temple (PTT), Taiwan’s largest online bulletin board where the issue of US pork has been extensively discussed. In their study of stance detection of comments posted on PTT, Chuang and Hsieh (2015) examine the cross-fertilization of ideas between netizens participating in online discussions and audiences reading the comments. Indirect participation in the conversations notwithstanding, the audiences still receive the messages conveyed by the interlocutors. Hence, discussion on PTT represents a dominant force shaping public attitudes toward a particular issue. However, studies to date on stance detection have frequently drawn on social media data at a document-level where each text was written by …

Content truncated to 500 words

Labeled Text

As the import of pork with ractopamine from the United States (US) has stimulated a heated discussion on social media, the purpose of this exploratory study is to assess public opinion on the import of US pork by examining the interplay between the two sub-issues, namely food security (FS) and political economy (PE), at the word, sentence, and time series levels. First, we collect comments related to the US pork issue on PTT, the largest online discussion forum in Taiwan, throughout the year 2021. Second, at the word level, we apply Latent Dirichlet Allocation (LDA) to extract the main topics. Third, at the sentence level, we sample a subset of data for manual annotation. Taking the annotated subset as the gold standard, we use the pretrained BERT-base-Chinese model for the downstream topic classification task to predict the whole dataset. Fourth, at the time series level, we employ Detrended Fluctuation Analysis (DFA) to evaluate the predicted number of comments for FS and PE sub-issues. Two main results are obtained: (1) the theme of political economy interweaves with the concept of food safety at both word and sentence levels; (2) DFA shows the absence of a more salient perspective between the FS and PE sub-issues at the time series level. 1

Publication Details

Paper ID:
28795
Year:
2022
Publication:
TIPSTER
Publication Type:
conf
DOI:
nan