Research on Patent Text Classification Based on Graph Neural Network
Traditional patent classification is carried out manually by experts.With the development of big data,artificial intelligence and natural language processing technology,automatic classification of patents is becoming one of the important research directions in both academia and industry.The text classification technology can be applied to determine whether a patent application can be granted,aiding in the automation of processing and analyzing a large number of patent documents,thereby improving work efficiency.This paper focuses on the English texts from a vast number of patents,and proposes an automatic patent text classification method based on the graph neural network model,which is used to assess whether patent applications can obtain authorization.This article utilizes the deep learning algorithm TextGCN to learn on patents'abstracts,leveraging the neighbor information and node features of graph-structured data.Through the neural network,it generates representation vectors for patents,facilitating the forecasting of patent authorization results.The experimental findings demonstrate that the deep learning approach applied in this study achieves commendable classification results.Compared to the Doc2vec and TFIDF methods,the TextGCN model shows improvements in terms of precision,recall,accuracy,and F1 score.This method can offer a dependable research foundation for automatic prediction of the grant status of patents.
万方数据
维普
