Digital news ecology and polarization in the context of deep learning: empowerment, entrenchment and reconciling failure

Abstract

This paper aims to discuss the causes of public opinion polarization in digital news ecology and then analyze group polarization’s specific manifestations and carriers. Secondly, it introduces word embedding technology to analyze digital news text, utilizes a multi-head self-attention mechanism to construct a new classification model, and realizes collaborative filtering recommendations of digital news based on users’ interests and news hotspots. Finally, social media is taken as an example to analyze the factors that cause polarization in digital news ecology. The results show that the digital news ecosystem can empower the development of the polarization phenomenon, deeply involve users, and group users by solidifying their identity construction, social endorsement and other directions. It can also provide a space for the polarization phenomenon to be reconciled, but due to the emotional loss of control and the decline of publicity leading to the failure of reconciliation, the similarity of sadness emotion reconciliation is 36.94%. It is suggested that the digital news ecology provides technical support for developing polarization phenomena, enabling them to thrive in digital news.