Sentiment analysis of college students’ microblog posts regarding digital identity via knowledge integration framework

This study investigates the digital identities and online sentiments of college students by leveraging the vast content they generated on social media platforms, specifically Sina Weibo. Utilizing the Enhanced Representation with Knowledge Integration Framework (ERNIE), we developed an optimized sentiment identification model that achieved a remarkable accuracy of 89.51%. Our findings reveal that 78.99% of Weibo posts express either positive or negative sentiments, with positive emotions predominantly surfacing in various factor analyses. Moreover, a strong negative correlation is identified between positive and negative sentiments. The sentiment analysis conducted in this study not only offers a novel perspective for understanding the digital identity of college students but also underscores the necessity for higher education institutions to provide effective educational guidance and management of digital emotions. An algorithmic model was also devised to track the sentiment trajectory of individual users, revealing that a significant subset (21%) exhibits negative sentiment trends. This underscores the imperative for targeted education and guidance in managing online emotions among college students. Notably, a minority of users demonstrated a consistent decrease in sentiment polarity, indicating an imperative for student affairs professionals to monitor and assist individuals exhibiting extreme sentiment expressions on social media.