Hot Public Appeal Extraction and Visual Analysis Combined BERT and Spatio-Temporal Location

Abstract

Analyzing the temporal and spatial distributions of public appeal can contribute to understanding issues of public concern. However, how to find hidden and valuable for early warning information and intelligence clues in large-scale text data remains a challenging task, especially with the explosion of public appeals. This paper proposes an intelligent extraction model of hot public appeal based on BERT and Spatio-Temporal location. For the unstructured appeals and its address, the fusion model is trained after extracting the keywords by word segmentation. Taking Baibuting Community as an example, experimental results show that the text representation model and text clustering method proposed in this paper have better topic detection effect than traditional methods. Combining spatial and temporal information with the classification results, We analyze the spatial and temporal distribution characteristics of hot topics in different regions. It will assist the grass-roots level in the rapid discovery, rapid response and coordinated disposal of the main problems of public appeal, and improve the refinement and intelligence level of grass-roots social governance.