2.2.3. Emotion Analysis Based on SnowNLP Model

Emotion analysis, the important work of opinion mining, aims to analyze people’s emotional situations towards entities such as topics, services, products, organizations, and their attributes [74,75]. Associated with social media data, automatic emotion analysis of the data is helpful to grasp public opinions in order to make corresponding decisions [76,77].

Bayesian classifiers SnowNLP (Available online: https://github.com/isnowfy/snownlp, accessed on 20 August 2022), which is the Scikit-learn Python machine learning library, have been extensively used to analyze the emotions of Chinese texts in research on public health. Zhang et al. used the SnowNLP model to analyze the sentiment data of restaurant reviews, and studied the differences in consumer preferences and emotional changes before and after COVID-19 [78]. Moreover, based on Weibo comment data, the SnowNLP model was used to analyze the emotional evolution of epidemy-related topics and visually display them in time and space [79]. Jia et al. collected Weibo public comments on returning to college after the epidemic and analyzed sentiments based on the SnowNLP method [80]. The results showed that students had positive emotions about going back to college. Zhang et al. used a Python script to collect 65,277 valid Weibo comments to assess the environmental risk perception of urban residents [81]. SnowNLP sentiment analysis was used to measure environmental risk perception in 366 cities in China, and the structural proportion characteristics and temporal and spatial differences of environmental risk perception were analyzed.

Based on the application experience of the above research, this paper also uses the SnowNLP model to analyze the consumers’ emotions towards different public health effects of online games and complete the last NLP task. For government and enterprises to grasp the opinion information related to online games’ domains, the emotion analysis task is specially designed, which is based on the results of multi-label classification, that is, to analyze the emotions from different consumer’s opinion regarding recognition results of online games.