Social Media Monitoring Of Involution Phenomena Based On Hidden Topic Markov Model And Sentiment Analysis

In recent years,the word ’involution’,which has been popular on the Internet,has lost its original meaning and now refers to the phenomenon that peers have to make more efforts to win limited resources,resulting in a decline in ROI.Since involution has brought increasingly negative impacts to society,it is of great significance to grasp the dynamics of public opinions about involution.However,there are few text analysis studies around the social hotspot of involution at present,most of which focus on involution phenomena in a specific field,and the data sources are quite single.Text analysis of reports and discussions can reflect the impact of involution in different fields.In this paper,news articles concerning involution on 36kr.com and the answers to the question ’How to get rid of involution in China?’ on zhihu.com are obtained through web crawling,which are taken as the corpora of news and folk speech respectively.After data cleaning,word segmentation and stop words removement,half of the texts are randomly sampled from the two corpus and sentiment labels are tagged manually.Based on the tagged datasets,sentiment analysis are conducted through BERT,Word2Vec+BiLSTM,BERT+Bi-LSTM respectively.As a result,the BERT+Bi-LSTM model proves to be best of the three,and the accuracy of the test datasets reaches 0.75.Then the optimal model is used to predict the untagged datasets,and it is concluded that both corpus tend to be negative,and folk speeches are more pessimistic,which reveals public opinions are one-sided.Next the hidden topic Markov model is used for topic modelling,and it is found that news topics can be summarized as three aspects:excessive competition of industry and brands,investment growth,and career advancement and development,while among netizens,more discussions are held around resource allocation,reform,market involution,education involution,career advancement and so on.It shows that the opinions of netizens are more flexible than those of standard-format news texts,and excessive competition of industry and brands is gaining growing concern from the media.Besides,the heat of tracking investment and growth has not diminished and the proportion of negative news has always maintained a high level.Finally,according to the coverage of involution and the means of eliminating the involution presented by the results,some suggestions including predicting consumers’demands,exploring differentiated markets,prioritizing technology and product innovation and so on are put forward for the industry.For individuals,we had better set up clear goals and help employees balance their work and life.