Technology Innovation Methods Based On Patent Citation Network Analysis And Text Mining Of Patents

Since the advent of 21 st century,with the rapid development of information technology,the pace of technological innovation is accelerating at an unprecedented rate.The proliferation of groundbreaking technologies,such as: big data,artificial intelligence,and Internet of Things,is fundamentally transforming how people live and work.The swift progression of technology has yielded an extensive volume of patent data.By leveraging these data,some innovation management methods can analyze technological trends,aid in the formulation of research and development strategies,and mine technology innovation directions.The application of these methods may ultimately enhance the technology innovation capabilities of individuals,corporates,and even nations.In addition,they contribute to the sustainable and rapid development of technology innovation.However,most existing methods consider patent citations and texts separately,which results in incomplete information usage and other issues when dealing with some innovation management tasks.Hence,investigating the combination of patent citations and texts in innovation management methods not only has the potential to overcome some of the limitations of existing approaches,but also provides a significant research perspective for innovation management field.The research scope of innovation management is wide-ranging.Technological trend analysis can provide valuable insights for researchers in making effective research and development strategies,as well as long-term planning,from a macroscopic perspective; Technology innovation direction mining can specifically assist in designing new technologies or products from a micro perspective.Both of them are core issues in innovation management,and have a decisive impact on the development of technologies and products.Hence,Therefore,this thesis proposes and delves into the following science issue:· How to integrate the technical association structures contained in patent citation networks with the technological semantic information contained in patent texts,in order to uncover the main technological development trends and mine crucial directions of technology in-novation?To address the issue,the following research works are carried out: 1)research on technological trends integrating patent citations and abstract semantics,2)research on technology fusion evolution integrating patent citations and semantic analysis of claims,3)research on the removal of low-quality citations and innovativeness trend based on the semantics of patent claims,4)research on technology innovation direction mining based on the syntactic dependency relationships of patent claim texts,5)research on technology innovation direction mining based on patent citations,claims,and deep learning.The main research contents and results are as follows:1)Research on technological trends integrating patent citations and abstract semanticsA method for exploring technological trends has been proposed by comprehensively considering patent abstracts and citations.Patents are categorized into different topics according to their research contents.Then a set of global and local importance indices are proposed by integrating citation relationships and abstract semantics.The former can effectively mine the most important patents in a technological field through considering the semantic information of patent citations and abstracts.The latter can reveal the most important technologies during various periods,and further be adopted to construct the main paths of a technology domain.According to these main paths,the development process of this domain can be deeply revealed.Experimental analysis has shown that the proposed methods can mine important technologies in a field accurately,and the main paths can uncover the development situation of multiple sub-technology fields in the domain better.2)Research on technology fusion evolution integrating patent citations and semantic analysis of claimsA method has been proposed to explore new research hotspots and reveal the process of technology fusion evolution through patent claims and citations.Based on the cosine similarities between patent claims,a technological inheritance indicator is proposed to detect important claims in different periods.Experimental analysis has shown that the index can take the semantic information in claims and technological inheritance relationships in citations into consideration effectively.Then,the important claims in different periods can be filtered more accurately.On this basis,the main paths can be constructed,and a method has been proposed to reveal the main technological origins of patents in the paths.Thereafter,the technology fusion evolution process can be uncovered.Experimental analysis has shown that the proposed method can reveal the diffusion and development principles of technology deeply.3)Research on the innovativeness trend and removal of low-quality citations based on the semantics of patent claimsA method has been proposed to analyze the trends in technological innovativeness and remove low-quality citations while taking into account the semantic information of patent claims.Two indicators have been proposed to uncover the changing process of patent innovativeness over time from local and global perspectives by term frequency-inverse document frequency(TF-IDF),patent claims,and citations.Experimental analysis has shown that technological innovativeness decreased gradually over time,and the approach is effective in revealing the technological trends from the perspective of innovativeness.Previous studies have demonstrated that the quality of citations has a significant impact on many citation-based analysis methods.Hence,removing low-quality citation relationships is crucial to improving the effectiveness of such methods.For dealing with this issue,an approach is devised based on the two indicators,to remove the citation relationships with low relevance from both local and global perspectives.Experimental analysis has demonstrated the suggested approach is effective in simplifying patent citation networks and retaining high-quality citations.In addition,it can enhance the efficacy of various citation network analysis methods indirectly.4)Research on technology innovation direction mining based on patent claims and syntactic dependency parsingBased on the patent claims of a technology domain and syntactic dependency parsing,a patent technical element dependency network(PTEDN)can be constructed,and some valuable usages have been proposed.By conducting syntactic dependency parsing on the patent claim texts,a network containing a large amount of text information has been proposed.This network can provide more comprehensive and detailed information.Experimental analysis has shown that compared with traditional knowledge graphs,the proposed network contains wordlevel information,and is with higher completeness and finer granularity.Based on a PTEDN,a heuristic algorithm has been proposed,which can perfect the ideas and assist in developing new technologies based on a researcher’s initial knowledge set and research interests.Experimental analysis has shown that compared to traditional knowledge graph-based methods,this approach can provide more detailed ideas,and the heuristic technology innovation direction mining appoach is more flexible.5)Research on technology innovation direction mining based on patent citations,claims,and deep learningTaking both patent citations and claims into consideration,a method for technology innovation direction mining has been proposed by using deep learning.Through the utilization of deep learning,a comprehensive understanding of the semantic information contained in patent claims can be obtained.Moreover,this understanding can be creatively integrated with the technological development implication of patent citations,leading to the revelation of technological development phenomena from the aspect of innovation perspective.Experimental analysis has shown that the proposed approach can effectively evaluate the similarity of patent innovation perspectives,and reveal the trend of patent innovation perepective similarity over time.Based on the similarity,an approach has been proposed to mine technology innovation directions by drawing upon historical development processes.It involves integrating the technological development implication of citations with innovators’ technologies.Experimental analysis has demonstrated that the proposed approach can mine technology innovation directions from a novel perspective,provide new research perspectives for the combination of deep learning and innovation management.This thesis addresses the current research situation in which most traditional innovation management methods consider the texts and citations of patents,separately.By fully exploring the combination of patent text semantics and citations,this study explores and solves the issues of technological trend analysis and technology innovation direction mining from a novel perspective.The research findings of this thesis provide valuable guidance for the future development of innovation management and real technology development.