Design And Implementation Of Innovation Service Engine Based On Patent Knowledge Engineering

Innovation is the key driving force for technological development and accelerating economic growth.The implementation of innovation-driven strategy has become a national strategy,providing new drivers for economic development.Patents are the most typical intellectual property rights,containing rich knowledge innovation.The mining and effective application of patent knowledge can promote the innovation development,assist companies and individuals in innovative design,and provide technical support for mass entrepreneurship and innovation.But patents exist in unstructured text,making it difficult for computers to understand the implicit innovative knowledge they contain,especially the purpose,mechanism and inventive principles of patents.How to mine valuable innovative knowledge from massive patent text data and clearly display the complex association information between patent data faces a huge challenge.The development of big data and artificial intelligence technologies has provided opportunities for patent innovation knowledge extraction and application Based on the support of national ministry of science and technology innovation method projects,this paper proposes to mine and analyze the implicit innovative knowledge contained in patent texts through knowledge engineering methods including knowledge graph and information extraction technology to assist and guide users' innovative activitiesThis paper designs and implements an innovation service engine based on knowledge engineering,providing services such as patent innovation knowledge extraction,patent citation relationship analysis,technology trend analysis and prediction,and cross-domain patent recommendation for enterprises undergoing technological upgrading and transformation,scientific researchers and general public users with innovative knowledge needs,fully address the innovative information needs of different users.The main functional modules of the system include patent structured information extraction and analysis,data dynamic management based on patent knowledge graph,convenient patent retrieval,patent information visualization and analysis,and patent recommendation based on innovative pattern.The patent structured information extraction and analysis module includes patent data acquisition,patent data pre-processing,neural network model construction,and analysis of structured information extraction results,displaying tacit knowledge such as purpose,mechanism and inventive principles of patents to users.The data dynamic management based on patent knowledge graph module can support large-scale patent document graph semantic processing.The patent information visualization and analysis module includes inventor map,invention-institution map,agency map,patent citation map and technical field map,etc.It can analyze patent-related information visually and help to understand and predict the development trend of technology.The convenient patent retrieval module provides retrieval methods based on different fields,including keyword-based retrieval,topic-based retrieval,inventor-based retrieval and IPC-based retrieval.Patent recommendation based on innovative pattern module includes the integration of innovative pattern and patent recommendation module.This module allows user to custom innovative pattern based on the learned "purpose,mechanism,invention principle" structured representation and can meet the different information demand of users.During the system development process,based on cutting-edge technologies such as entity extraction and deep learning,we propose a neural network model to extract structured information including purpose,mechanism and inventive principles and improve the accuracy of knowledge extraction results and provide technical support for patent recommendation based on innovative pattern.We use the graph database Neo4j to store patent knowledge maps,provide a set of data operation engines to add,delete,modify and check patent data,and support patent information reasoning and innovative patent recommendation services based on the semantic relationship between entities.We use the Echarts visualization framework to achieve multi-dimensional correlation analysis of patent data.This paper makes a comprehensive description of the innovation service engine from the aspects of requirements positioning,summary analysis,detailed design and implementation,and system testing.The innovation service engine has realized the collection and management of more than 3 million patents.It uses a neural network model to extract and analyze the hidden innovation knowledge such as the innovation purpose,innovation mechanism,and principle of the invention in the patent text with extraction accuracy of 93.5%,which can meet the users' information demand for patent innovation knowledge.The innovation service engine has been demonstrated in many innovation method training institutions and technology transfer centers,providing patent value analysis,technology trend analysis and other services to more than 2,000 users,assisting users to summarize more than 10,000 creative solutions and technological transformation solutions,which stimulates users' innovation ability and improves the efficiency of innovation activities.