Natural Language Processing in assistance to Inventive Design activities

• Published in: Procedia CIRP Vol. 109; pp. 7 - 12
• Main Authors: Berdyugina, Daria, Cavallucci, Denis
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 2022
• Subjects: Augmented creativity; Inventive Design Method; Natural Language Processing; TRIZ; TRIZ; Augmented creativity; Inventive Design Method; Natural Language Processing
• ISSN: 2212-8271; 2212-8271
• DOI: 10.1016/j.procir.2022.05.206

Design activity requires engineers to use their knowledge to understand, formulate and solve industrial problems whose complexity is constantly increasing. Our work, at the frontier of linguistics, computer science and engineering sciences, consists in making optimal use of the information contained in texts useful to designers, particularly patents. Existing method to extract information from patents already exist, but without taking into consideration the use of their output as a mean to populate Inventive Design ontology. This paper describes how we use Natural Language Processing to extract information useful to designers in the context of an inventive activity. Our ontology inspired by the Theory of Inventive Problem Solving (TRIZ) helps us to categorize the information collected in corpus consisting of thousands of unstructured texts, to redistribute it to the designer and thus increase his ability to formulate and solve the problems that his subject of study confronts him with. This study concerns a new way of categorizing patents into newly defined technical domains so as to better perceived a prior art during the early stages of an inventive process. Our results attest significant progresses both in terms of the speed to understand data complexity and in terms of the accuracy of the perception of an explored field prior to enter into a resolution logic.