The Mechanism and Application Prospects of Transformer Model in Emotional Analysis of Film and Television Reviews

Liu Heming

Jinan New Channel -JUTES High School

Abstract：

This paper takes the Transformer model as the core, systematically explaining its self-attention mechanism, multi-head attention, position encoding, and encoder-decoder architecture in the theoretical adaptability and operational mechanism of sentiment analysis in film and television reviews. The study shows that the Transformer overcomes the information attenuation problem of traditional models through global dependency capture, and multi-head attention realizes the parallel extraction of multi-dimensional sentiment features. This paper discusses the theoretical path, such as fine-grained emotional linear evolution, pre-training and fine-tuning knowledge transfer, and cross-modal integration. He also discussed their value to the industry and what might happen in the future. We look at this from the perspectives of founder's feedback, platform decision-making and public-oriented service.

Key Words：

transformer model; film and television reviews; sentiment analysis