Patents are one of the main innovation knowledge sources for engineers and companies. Inventive Design Method (IDM) - results from a research that extends from TRIZ and contains formal knowledge description components using ontologies, such as problems, partial solutions, and parameters. In this paper, we introduce IDM-Similar model that extends existing research work in IDM-related knowledge. A neural network named Word2vec and cosine similarity approach are used to build this model to compute the similarity among problems in wide range domains' patents covering from the chemistry to mechanics and the computer to physics. Our model assumes that a partial solution of a patent could be used to solve the problem of another patent from a different domain if these two problems are similar enough. Experiments show that our model is a promising alternative to classical TRIZ for engineers to associate their problems in a field to solutions from patents of another field. Consequently, the step dedicated to solution concepts ideation is improved using our work.
展开▼
