以人工智慧分析香港來台移民潮之媒體框架及留言

Abstract

自2019年起，香港人移居台灣的人數顯著增加，成為台灣社會與媒體關注的重要議題。香港移民潮不僅反映兩岸三地間的政治關係，也成為台灣內部對移民與中國議題態度的觀察指標。而媒體如何建構議題，或對社會輿論具有潛移默化的影響。有鑑於此，本研究旨在探討台灣立場不同的媒體如何報導香港來台移民議題，以及報導下方留言如何反映民意與媒體效應。本研究以台灣十個主要新聞媒體的YouTube頻道作為資料來源，採用BERTopic主題建模與BERT文本分類等人工智慧技術，分析報導影片的媒體框架、網民留言主題，以及兩者之間的關聯。研究發現，不同政治立場的媒體在報導數量與框架使用上存在顯著差異。泛綠媒體相較於泛藍媒體，傾向更積極報導香港移民議題，並多採用「多元文化」框架，強調民主自由與文化多樣性；而泛藍媒體則偏好使用「經濟」框架，聚焦移民帶來的實際影響與潛在風險。而針對相關影片下的留言，透過BERTopic主題建模，發現討論主題以兩岸關係及政治為主，其次才是實際影響的討論，如房價、醫療和移民門檻等，顯示香港移民議題重要性在政治上最為顯著。本研究進一步分析媒體框架與留言的議題屬性、情感及立場之間的關聯後發現，影片使用不同媒體框架，留言中的議題屬性、情感與立場皆有差異，如採用「多元文化」框架，會較少負面情緒留言，採用「經濟」框架下的留言，則抱有較強的反移民立場。本研究旨在探討對香港來台移民的媒體框架與民眾態度，填補台灣民眾對香港移民態度研究之不足。方法上採用大數據與人工智慧技術，務求為移民研究開拓新路。Since 2019, the number of Hong Kong residents relocating to Taiwan has increased significantly, making it a prominent issue in both Taiwanese society and media. The wave of Hong Kong immigration not only reflects political relationship between Taiwan, Hong Kong, and mainland China but also serves as an indicator of Taiwan’s internal attitudes toward immigration and China-related issues. Media framing of such topics can subtly influence public opinion. In light of this, this study aims to examine how Taiwanese news media with different political orientations report on the Hong Kong-to-Taiwan immigration issue, and how comments under these reports reflect public sentiment and media effects.Using YouTube channels of ten major Taiwanese news outlets as data sources, this research employs artificial intelligence techniques such as BERTopic topic modeling and BERT-based text classification to analyze the media frames used in the reports, the themes of online user comments, and the relationships between them.The findings reveal significant differences in the frequency and framing of reports depending on the political stance of the media. Compared to blue-leaning media, green-leaning outlets are more active in reporting on Hong Kong immigration and tend to adopt a"multicultural" frame, emphasizing democracy, freedom, and cultural diversity. In contrast, blue-leaning media prefer an "economic" frame, focusing on the practical impacts and potential risks posed by immigration. Regarding the comments under related videos, topic modeling with BERTopic shows that most discussions center around cross-strait relations and politics, followed by concerns about practical issues such as housing prices, healthcare, and immigration thresholds. This indicates that thepolitical significance of the Hong Kong immigration issue outweighs its practical considerations.Further analysis of the relationship between media frames and the thematic attributes, emotions, and stances in the comments reveals that different frames lead to varying patterns of public response. For instance, the"multicultural" frame tends to attract fewer negative emotion comments, while the "economic" frame is associated with stronger anti-immigration sentiment in the comments. This study explores the media framing and public attitudes toward Hong Kong immigrants in Taiwan, addressing a gap in the existing literature on Taiwanese public opinion regarding this group. Methodologically, it leverages big data and artificial intelligence technologies to open new avenues for immigration research.