分析文本情感是自然語言處理中一項常見的任務，在多年發展後已結合各另領域 應用於多項任務上，其中，電子口碑（Electronic Words of Mouth, eWoM）透過收集消 費者評論來分析消費者的情緒波動，進而幫助企業可以改善產品服務。在過去，消費 者評論的相關研究並不會探討輸入硬體設備造成的影響，然而近年有論文指出硬體設 備會影響英語使用者所傳達的情感強烈程度。 本文參考過去論文方法由 Tripadvisor 與中文社群平台巴哈姆特獲取新的文本資料， 搭配上 ANTUSD 情感詞典計算文本中情感詞頻率、情感詞強度，以及 VADER 模型分 析文本中句子的情感強度，最後以 Man-Whitney U 檢定及 Rank-biserial correlation 效果 值分析數值結果，來探討中文使用者是否受到輸入設備影響情緒表達強度，最終結果 發現在本研究設計的多個實驗中呈現了統計上顯著而效果值低的結果，在少部分的實 驗檢測到具有顯著性且有效果值存在的結果，透過實驗成果本研究發現到在不同網站 所收集到的資料並沒有一致性結果，情緒強度雖然確實因硬體設備有差異，卻並非是 手機群體所產生的強度必定大於電腦群體。;Analyzing sentiment is a common task in natural language processing, which has been applied to various domains and tasks over the years. One such domain is electronic word of mouth (eWoM), where consumer reviews are collected and analyzed to understand consumer sentiment and help improve products and services. Previous studies on consumer reviews rarely explored the impact of hardware devices on the expressed emotions. However, recent research has indicated that hardware devices can influence the intensity of emotions conveyed by English users. In this study, we collected new textual data from Tripadvisor and the Chinese community platform Bahamut, and used the ANTUSD sentiment lexicon to calculate the frequency and intensity of sentiment words in the text. Additionally, we used the VADER model to analyze the sentiment intensity of sentences. The results were analyzed using the Mann-Whitney U test and the Rank-biserial correlation effect size to investigate whether Chinese users are influenced by their input devices in terms of emotional expression intensity. The findings of this study indicate statistically significant but practically low effect sizes in multiple experiments conducted within the research design, only in few experiments found the statistically significant and effect sizes exist. Based on the experimental findings, this study discovered that the data collected from different websites did not yield consistent results. Although there was indeed variation in emotional intensity due to hardware devices, it was not necessarily the case that the intensity generated by the mobile device group was always greater than that of the computer device group.
