<<
A patent was granted for a base algorithm for text-based recommendation.
2023 December 7
We have patented a document analysis algorithm for understanding qualitative information such as user's tastes from text.
This invention does not require the accumulation of user behavior data, but analyzes from text the tastes and personality characteristics of the user who described it. We've created a new way of machine learning algorithm which learns multi-semantics in a high-dimensional word embedding space according to any user defined concepts and it efficiently computes the qualitative information implied in a text in a fraction of the computation time.
It not only solves the cold-start problem of conventional recommendation (i.e., it does not work unless data is accumulated), but also enables a more precise understanding of diverse user needs.
Document No.: Patent 7393772
Application No.: Patent Application 2022-180651
Inventor: Kazuki Otsuka
Publication Date: December 7, 2023
This invention does not require the accumulation of user behavior data, but analyzes from text the tastes and personality characteristics of the user who described it. We've created a new way of machine learning algorithm which learns multi-semantics in a high-dimensional word embedding space according to any user defined concepts and it efficiently computes the qualitative information implied in a text in a fraction of the computation time.
It not only solves the cold-start problem of conventional recommendation (i.e., it does not work unless data is accumulated), but also enables a more precise understanding of diverse user needs.
Document No.: Patent 7393772
Application No.: Patent Application 2022-180651
Inventor: Kazuki Otsuka
Publication Date: December 7, 2023