Digital media recommendation system design based on user behavior analysis and emotional feature extraction

Abstract

In the contemporary digital era, multimedia platforms, such as social media, online comment sections, and forums, have emerged as the primary arenas wherein users articulate their sentiments and viewpoints. The copious volume of textual data generated by these platforms harbors a wealth of emotional insights, which are paramount in comprehending user behaviors, fine-tuning content dissemination strategies, and elevating user satisfaction. This scholarly paper introduces an innovative framework, denominated ATLSTM-PS, for formulating content dissemination strategies within digital media platforms predicated upon a user-centric emotional perspective. Initially, it accomplishes extracting emotional content from users’ commentaries on digital media platforms, amalgamating the ATT-LSTM method with the attention mechanism, resulting in enhanced feature extraction precision compared to traditional single RNN and LSTM approaches. Subsequently, the framework extracts information at the feature layer by integrating user behavioral and emotional attributes. Following this, by amalgamating user behavioral and emotional features, ATLSTM-PS affects the synthesis of feature layer information. This meticulous amalgamation yields highly precise recommendations that cater to user demand. Empirical results obtained from publicly available and proprietary datasets substantiate that ATLSTM-PS substantially enhances the efficacy of content dissemination through the synergy of distinct attention layers. This research contributes not only a novel technical tool in sentiment analysis but also furnishes a potent methodology for multimedia platforms to refine their information dissemination strategies, thereby augmenting the user experience.