Setting Up Context-Sensitive Real-Time Contradiction Matrix of a Given Field Using Unstructured Texts of Patent Contents and Natural Language Processing

摘要

It is well known that Altshuller matrix is the most frequently used tool by TRIZ practitioners. While experts often turn away from it in favor of more recent (and reputedly more effective) tools such as Vepoles and ARIZ85C, it is clear that beginners prefer the matrix because of its simplicity. Nevertheless, two sensitive phases in its use call it into question. The association of the user's specific problem with one of the 39 generic parameters that listed in matrix and the interpretation that can be made of the inventive principles proposed to users. We have developed an approach based on Natural Language Processing to process a specific corpus of patents corresponding to a given technical field in real time. This leads to a new approach developed in this article to propose to users a new matrix for each study but which considers the vocabulary of a given domain to describe the oppositions between parameters. Such a matrix could constitute a state of the art in form of contradictions of the field being explored, which we believe will ease its use upstream of inventive studies processes in order to target the resolution of key and up-to-date contradictions of the same field.