Neural semantic evaluation of blockchain policy tools in China

To explore the differences in policies guiding the development of the blockchain industry in different regions of mainland China, in this study, we developed a compound neural network model comprising bidirectional long short term memory and deep biaffine attention models that analyses the semantic texts of blockchain policies issued by 31 provinces in mainland China. Machine learning models — specifically, term frequency–inverse document frequency and K-means models are used to implement feature selection of the policy text matrix classification after semantic analysis. Finally, this study proposes an innovative policy tools matching approach. We construct a word-topological map for each text category based on semantic relationships. To validate the effectiveness of these tools, we conduct an empirical analysis using a multivariate linear regression model. The results demonstrate that blockchain policy tools significantly promote blockchain innovation in Mainland China.