End-to-End Multimodal COVID-19 Content Quantitative Safety Detection Algorithm

Abstract

This paper takes the multimodal COVID-19 related dataset as the research object and carries out research on the content safety monitoring algorithm in the field of COVID-19. It is proposed to carry out research work in the following two aspects. The first is end-to-end topic and sentiment modeling. Turning topic modeling and sentiment analysis into multi-label classification problems essentially unifies topic modeling and sentiment analysis, with the model directly outputting the predicted sample's topic and sentiment. The second is to quantify sentiment analysis results. While modeling sentiment polarity, it also models sentiment intensity. By analyzing public opinion through sentiment polarity and judging people's attention to a topic and the degree of speech safety based on sentiment intensity values, we propose an end-to-end multimodal COVID-19 content quantitative safety detection algorithm (EMQD). The EMQD algorithm is tested on the KJZY2020 dataset and compared with other algorithms.