使用主題模型技術探索應用程式評論

Abstract

在大數據時代，網路的普及導致了用戶生成內容的激增，包括應用程式評論。這種數據的激增既帶來了機會也帶來了挑戰，用於提取有意義的見解。本研究旨在利用主題模型技術，將大量的應用程式評論數據轉化為可操作的見解。通過應用主題模型方法，我們試圖識別評論中的潛在主題和趨勢。我們使用以 BERTopic 為基礎的方法，並比較其他主題模型技術來分析評論數據。我們的數據集來自熱門應用 TikTok 的評論，這些評論提供了多樣且全面的用戶意見和反饋。為了評估主題模型的性能，我們採用了多種評估指標，包括連貫性、多樣性和可解釋性，以確保模型生成有意義且有用的主題。我們將各種不同的結果以多元的可視化技術來展現，有助於有效地傳達從數據中獲得的見解，使得結果更容易理解和解釋。

In the era of big data, the proliferation of the internet has resulted in an overwhelming amount of user-generated content, including app reviews. This surge in data presents both opportunities and challenges for extracting meaningful insights. This study aims to harness the power of topic modeling techniques to transform vast amounts of app review data into actionable insights. By applying the topic modeling method, we seek to identify underlying themes and trends within the reviews. We utilized methods based on BERTopic, including comparisons with other topic modeling techniques, to analyze the review data. Our dataset consists of reviews from the popular app TikTok, which provides a diverse and comprehensive collection of user opinions and feedback. To assess the performance of the topic models, we employed several evaluation metrics. These included measures of coherence, diversity, and interpretability to ensure the models generated meaningful and useful topics. We represented the various results using diverse visualization techniques, which help in effectively communicating the insights derived from the data, making it easier to understand and interpret the findings.