Innovative Sentiment Analysis of Mobile App Review Texts Based on MHA-MLP-Mixer

The current landscape of the mobile application market unfolds as a rich and diverse tapestry, with developers struggling to meet user demands and distinguish themselves amidst fierce competition; users confront a vast array of applications, each with unique features, posing challenges in accurately aligning with their actual needs. The star rating system, a commonly used evaluation tool in the industry, lacks in-depth interpretation of user comments, leaving developers unable to discern specific user needs and users deprived of nuanced reference information. Despite its crucial role as a key tool for deciphering emotional information in user comments, sentiment analysis technology still grapples with challenges such as its limited capability in extracting emotional features from a singular perspective and inadequate modeling of complex emotional interactions. Hence, this paper introduces a novel deep learning model, the MHA-MLP-Mixer model, specifically designed for sentiment analysis of mobile application review texts. Its innovation lies in integrating local and global text information, focusing on multi-perspective sentiment cues using a multi-head attention mechanism, and modeling deep sentiment interactions through MLP-Mixer. Experimental results on publicly available datasets demonstrate that the precision, recall, and F1 score of the model all exceed $\mathbf{8 5 \%}$, surpassing the analysis results of other models. This framework contributes to fostering the development of the mobile application market and enhancing user experience.