Development Of Robot Design Question Answering System Based On Patent Knowledge Graph

In recent years,intelligent manufacturing represented by robots has set off a worldwide storm in the field of production technology.The major breakthroughs in information technology,such as the Internet of Things,big data,cloud computing,artificial intelligence,etc.,have made the application of robots extend continuously from industrial manufacturing to healthcare,military,agriculture and other fields.The "Made in China 2025" Project,which is known as the Chinese version of "Industry 4.0",lists the robot industry among the key strategies.However,because the robot industry covers machinery,electronics,sensor detection,computer,life science and other disciplines,it puts forward higher requirements on robot design.At present,question answering system based on knowledge graph has been used in many specific fields,such as medical treatment,agricultural,and education.This thesis introduces the question answering system into the field of robot design,and proposes a question answering system for robot design based on patent knowledge graph.This question answering system can answer the questions in natural language on the aspect of robot design raised by the users,and provide the solutions and recommend the patents to the designers and researchers when they are learning and designing robots.The robot design question answering system based on the patent knowledge graph is developed based on the patent knowledge graph in the field of robots.At first,we use the methods of question template and question parsing to convert users' questions into query templates or question query graphs,and then expand the concept of entity in user question from the semantic perspective according to the third-party concept knowledge base and word vector model.Finally,the expanded query template or query graph of questions and patent knowledge graph are combined to establish the query matching relationship.The contributions of this thesis are as follows:(1)Unstructured user query questions in natural language are converted to structured representation.For user query questions,sentence similarity measure is used for template matching and constructing question query template.The Stanford Core NLP tool is also used to parse and build the question query graph for structuring achieve query questions.(2)The question sentences are expanded to deal with natural language ambiguity.The thirdparty concept knowledge base,e.g.,Concept Net and Microsoft Concept Graph,and the local patent data word vector model are used to expand the synonyms and related words of the entities in the structured query,and some filtering strategies are adopted.(3)Query matching is conducted for the expanded structured query questions and the patent knowledge graph.Exact matching and fuzzy matching are both used for the question query template and question query graph,and the relationship missing issue in the questions is also handled.(4)Robot design question answering system based on patent knowledge graph is designed and implemented.The patent knowledge graph is stored by using the Neo4 j graph database;different parts of the system is designed in a modular manner;the Django web framework is used for background processing,user interface design,etc.;and the j Query framework is used to complete the user interaction design.