Research on Migrant Works' Concern Recognition and Emotion Analysis Based on Web Text Data

Abstract

Based on the characteristics of convenience, autonomy, and equality, online self-media has become an important way for contemporary migrant workers to observe the world, understand society, examine themselves and express their demands. On the basis of the analysis of the domestic migrant works' concerns and their emotion analysis, we crawl data on Weibo about migrant works' topics as the basic corpus of migrant works' concerns, and then uses a combination of TF-IDF and Word2Vec methods to construct a recognition model of migrant workers' concerns. We found that wages, children's education, medical care and returning home are the main concerns of migrant workers. Meanwhile, further emotion analysis of the migrant works' concerns of using a deep learning model fused with Bi-LSTM and CNN was conducted. The results show that the proportion of negative emotion such as worries, complaints and impetuosity was significantly higher than that of other positive and neutral emotion like encourage and comfort. And the time when the negative emotion are concentrated is significantly related to the social events that occur in the corresponding time period. On the one hand, it shows that the concerns and emotion of migrant workers can be effectively observed and predicted through web text data. On the other hand, it also shows that the core well-being issues of migrant workers in the process of urban integration have not been effectively solved, and the government and relevant departments need to take targeted measures and give priority attention.