Conceptual Semantic Analysis of Patents and Scientific Publications Based on TRIZ Tools

Abstract

Being committed to the idea that problems from completely different fields could have conceptually similar solutions, Altshuller has analyzed more than 40,000 patents to identify and interplay those common inventive principles. However, reliable extraction and identification of different solutions among millions of patents and scientific publications is still a challenge. Inspired by the core notion behind the TRIZ, we have decided to build upon it by exploiting modern-day advances in both processing power and software engineering. In particular, the idea behind our methodology is to extract semantic features from large amount of source documents and subsequently subject them to analysis which is based on building the semantic boxes of their underlying concepts. Being created and identified, the semantic box allows extracting out-of-the-box ideas from different fields. To this day, our analysis has proved successful in processing 8 million patents and scientific publications with the use of machine learning and natural language-based processing techniques.