基于神经网络词嵌入Word2vec的舆情监测比较研究

Public opinion monitoring is a key information management tool that provides important data support to organizations by timely understanding social public opinion, helping them make flexible decisions, maintain reputation, and warn of crises, thereby effectively responding to changing social environments. This article uses the Word2vec method to compare public opinion monitoring with academic platforms and social Q&A platforms in China, aiming to promote research and practice in public opinion monitoring. Firstly, data was collected from both Zhihu and CNKI platforms to form corpora for social Q&A and academic platforms. Then, two Word2vec models were trained using these two corpora. Next, a comparison was made between social Q&A platforms and academic platforms based on the top 20 words that are most similar to public opinion monitoring. On this basis, the t-SNE algorithm was used to reduce the dimensionality of word vectors, and the matplotlib.pyplot package was used to visualize and compare the word embeddings of all words in two Word2vec models. The research results indicate that there is a direct difference in public opinion monitoring between academic platforms and social Q&A platforms in China. This study innovatively utilizes the Word2vec method and compares public opinion monitoring with academic and social Q&A platforms in China, providing a new perspective for public opinion monitoring research.