The Topic Evolution of Danmaku Text Based on BTM -- Taking the Context of COVID-19 as an Example
Xin Chen, Yixin Zhang, Junchao Wu, Lingyu Guo, Jiaxuan Chen, Jing Yang · 2022
The danmaku text is a popular interactive mode of instantaneous video comment. Websites with danmaku texts have been widely used recently, and this vast number of texts can be regarded as a short text mining resource. This paper uses Perplexity and Rényi entropy to evaluate the BTM (Biterm Topic Model), by extracting the topic from the danmaku texts to explore the evolution of danmaku text topics in videos relevant to COVID-19. The results show that Rényi entropy is an effective way to decide the optimal number of topics, and the topics captured by BTM indicate that video viewers showed positive attitudes in the face of this public health emergency.