Research On The Construction And Application Of Patent Knowledge Graph In Aerospace Field Based On Deep Learning

With the deepening of the country’s emphasis on the protection and application of intellectual property rights,it is more difficult to analyze high-quality patents in the face of the massive growth trend of patent texts.Aeronautics and astronautics is one of hightech industries of our country,and it contains a lot of high-tech achievements in the patent texts.It is an important foundation for scientific and technological innovation.Therefore,how to excavate and utilize the patent texts has become a hot topic of research in recent years.Due to the unique nature of patent text itself,the content information of the text is difficult to understand,and there are many proper nouns,so it is difficult for nonprofessionals to read and use it,especially in the field of aerospace patent mining and exploration research.Knowledge graph can show the complex relationship between text data in the form of graph,and further use these data for subsequent work.Deep neural network model has strong text semantic understanding and analysis ability,and can fully extract text features.Moreover,the model has fast processing speed,which can solve the problems such as inaccurate recognition and slow processing speed in the construction of knowledge graph by traditional methods,saving a lot of time and cost.Based on the above background,this paper uses the deep learning method to build the patent knowledge graph in the field of aerospace.Firstly,relevant patent data are retrieved from the patent database,sorted and cleaned the patent abstract data,and Word2 vec was used to vector represented the input text data.BiLSTM-CRF model was built to identify the named entity of patent texts in the field of aerospace,and was compared with BiGRU-CRF,LSTM-CRF and CRF models.The results showed that BiLSTM-CRF model had the best effect.The values of P,R and F1 are 83.37%,83.18%and 83.25%,respectively.Secondly,relationship extraction was carried out based on BiLSTM model.BiLSTM neural network is used to extract the feature of context timing information of target word vector,and Softmax classifier is used to output the result.The results show that the BiLSTM model has the best effect,and the values of P,R and F1 are75.50%,72.67% and 73.83%,respectively,which verifies the effectiveness of the constructed relational extraction model.Finally,the extracted patent relationships are reorganized and stored in the Neo4 j graph database,and the patent knowledge graph in aerospace field is constructed,and the question and answer application and visual display are carried out according to user requirements.In this paper,the patent knowledge graph in the aerospace field is constructed based on deep learning,which helps to better understand the key information of patents in the aerospace field and the relationship between them,improve the efficiency of patent text analysis in the aerospace field,facilitate patent staff to carry out patent management in this field,and promote knowledge discovery and trend prediction of advanced technologies in the aerospace field.To enhance intellectual property management capabilities in the aerospace sector and the competitiveness of the manufacturing sector,and promote innovative development in the aerospace field.