MBCA:Identification of high-value patents using deep learning based language understanding

Aiming at the problem that the existing high-value patent recognition methods rely on experts' subjective experience judgment and the patent recognition model cannot fully extract the feature information of the patent text, which leads to the unsatisfactory recognition efficiency of high-value patents, we propose a high-value patent identification method using deep learning based language understanding:MBCA.The experimental results in the chinese text classification dataset from Fudan University show that the accuracy of this method has improved by 1.59% compared with the optimal baseline method, and the empirical results in chinese patent dataset of lithography field show that the accuracy of this method has improved by 3.42% compared with the optimal baseline method, indicating that this method can fully improve the identification performance of high-value patents, which can provide an effective way for the intelligent identification of high-value patents and promote the effective transformation of patents, and has practical application value.