Evaluation of Public Participation and Emotion Classification for the Reconstruction of Old Communities Based on Semantic Analysis Algorithm

This paper delves into the assessment of public engagement and the categorization of emotional states to gain deeper insights into individuals’ involvement in the revitalization of historical communities and their emotional experiences during this process. First, this paper proposes an enhanced Faster R-CNN model for detecting and analyzing public participation. Specifically, our model incorporates attention mechanisms, both in the spatial and feature channels, to enhance the identification of human subjects and facilitate more precise target localization. Additionally, this paper addresses the challenge of detecting multi-scale targets by employing Feature Pyramid Network (FPN) technology, which enriches the features computed by Faster R-CNN. Furthermore, a method for describing public emotions based on semantic analysis is proposed. Leveraging semantic analysis algorithms, double LSTM and self-attention mechanisms are employed to categorize the emotions of individuals and generate captions to understand the emotions. Experiments validate the effectiveness of our approach in accurately pinpointing individuals within images, enabling a comprehensive assessment of public participation in the restoration of historic residential areas. Our method obtains a CIDEr value of 127.0 and an F value of 0.8856. Moreover, our method proficiently characterizes participants’ emotional states, providing valuable technical support for societal development.