Research on Topic Co-Occurrence Network and Evolution Path Based on DeepTFIDF Algorithm

This study proposes a keyword analysis algorithm, DeepTFIDF, based on deep learning methods. The algorithm constructs a co-occurrence network of key terms to visualize the evolution path of research topics over time. DeepTFIDF algorithm innovatively incorporates deep learning techniques by using word embedding models to map each vocabulary into a low-dimensional vector space, effectively capturing the semantic relationships between words. Additionally, the algorithm integrates the calculation of TF (term frequency) and IDF (inverse document frequency) weights, enhancing the accuracy of keyword selection. Experimental results demonstrate the effectiveness of the DeepTFIDF algorithm. Compared to traditional TF-IDF methods, DeepTFIDF exhibits better accuracy and interpretability in keyword selection and weight calculation. Lastly, this paper applies the DeepTFIDF algorithm to analyze the project data of the National Social Science Fund, constructing a co-occurrence network of key terms and visualizing the evolution path of research topics over time. The application results on the project data of the National Social Science Fund demonstrate that the DeepTFIDF algorithm effectively reveals the evolution path of topics and provides new perspectives on the trends of research topics.