A Property-Function based Approach to Patent Mining for Technology Foresight

Abstract

Modern economies emphasize the role of research and development (R&D) that promotes the creation, diffusion and accumulation of intellectual properties within economic systems. Technology foresight allows creating potential and intangible needs of customers by developing value-creating technologies or products for market integration and new market creation. Some previous approaches for technology trend analysis and technology evolution analysis are considered to be useful for technology foresight. However, two points should be complemented for experts to more focus on their knowledge services in that the approaches are relies heavily on knowledge and skills of experts such as domain experts or TRIZ (Russian acronym: the Theory of Inventive Problem Solving) experts, which are unavailable or expensive. They are 1) the dependency on domain experts’ knowledge in analyzing technological trends of emerging or related technology areas, and 2) dependency on TRIZ experts’ skills in identifying technology evolution patterns and their phases from given technologies. As a remedy, this research suggests a property-function based patent mining approach to address the problems. The property addresses ‘what a system is or has’, and expresses a specific characteristic of a system or its sub-systems

the function addresses ‘what a system does or undergoes’, and expresses a useful action of a system or its sub-systems. Properties and functions constitute the abstraction of a system, so they show directions in technological innovations of systems. The property is described mainly using adjectives, and the function is described mainly using verbs. Therefore information concerning properties and functions can be identified from technical information using Natural Language Processing (NLP) because they can be extracted by grammatical analysis. Based on the property-function based approach, this research performs three issues. They are 1) Invention Property-Function Network Analysis of Patents, 2) Mapping Patent Properties and Functions on TRIZ Trends, and 3) An Automated Method for Indentifying TRIZ Trends from Patents.The first issue, ‘Invention Property-Function Network Analysis of Patents’, deals with how to identify trends in technological innovation from new patents. The proposed method automatically extracts properties and function from patent text by exploiting NLP. Using properties and functions as nodes, and co-occurrences as links, an invention property-function network (IPFN) can be generated. The method identifies new technological trends from the IPFN by social network analysis (SNA) indicators including degree, centrality and density. The second issue, ‘Mapping Patent Properties and Functions on TRIZ Trends’, deals with a method to quantify and formalize the processes of TRIZ trend analysis. The TRIZ trend analysis can be used to support the design process and the managerial decision making process to predict further improvements of systems. The third issue, ‘An Automated Method for Identifying TRIZ Trends from Patents’, deals with how to automatically identify TRIZ trends from patents. As a study to facilitate the second issue, the third issue presents a method that consists of 1) extracting properties and functions in the form of binary relation from patent text by exploiting NLP, 2) defining a ‘reasons for jumps’ rule base that arranges trend specific binary relations for TRIZ trend identification, and 3) determining specific trends and trend phases by measuring semantic sentence similarity between binary relations in patents and ones in the rule base.Through these issues, this research provides following contributions: 1) providing methods for systematic identification of trends in technological innovation, 2) facilitating technology foresight for product design and managerial decision making, 3) supporting experts to more concentrate on their knowledge services, and 4) providing a basis for implementing a technology intelligence system.