As an important channel for governments at all levels to serve the public, the affective tendency of online reviews of mobile government APPs can have an important impact on users' online government satisfaction. In order to conduct fine-grained sentiment analysis of current mobile government APP user reviews, the article conducts sentiment tendency measurement of mobile government user reviews based on ABSA method:the LDA is used to extract implicit aspectual topics, and the expectation confirmation theory and information system success model are combined to construct a mobile government user demand model; meanwhile, the BERT is selected to conduct sentiment tendency probability for entity words At the same time, the BERT model is selected to determine the sentiment tendency probability of entity words, and then aspect-level sentiment intensity measurement is achieved through sentiment tendency matching among multiple dimensions. It is found that the method can explore the demand focus and emotional state of mobile government users, and the current users present positive emotional state for service quality, while the emotional evaluation of system quality and information quality is more negative.