Exploitation of Causal Relation for Automatic Extraction of Contradiction from a Domain-Restricted Patent Corpus

Altshuller contradiction matrix is one of the most popular tools among TRIZ practitioners, especially beginners, due to its simplicity and intuitive design. However, scientific and technological progress induces the constant appearance of new scientific vocabulary, which lower accuracy when using this static tool from the end of the sixties. Some attempts to rebuild the matrix or update it has been made within the past four decades but without any successful legitimation due to the lack of scientific proof regarding its relevance. Our recent findings in the use of Natural Language Processing (NLP) techniques allow the creation of a methodology for automatic extraction of the necessary information for establishing a domain-restricted contradiction matrix. In this paper, we relate a technique that exploits the internal language semantic structure to mine the causal relation between terms in patent texts. Moreover, the subject or domain restriction for a patent collection allows observing the links between extracted information at the over-text level. Such an approach relies on inter-and extra-textual features and permits a real-time extraction of contradictory relations between elements. These extracted elements could be presented in matrix form, inspired by The Altshuller contradiction matrix. We postulate that such a representation allows the construction of a state of the art in each domain, which will facilitate the use of TRIZ to solve contradictions within it.