Shotaro Kohama

Title: Shotaro Kohama and Automatic Ontology Generation

Introduction

Shotaro Kohama is an inventor associated with Palo Alto, California, and Mercari, Inc. His patent work includes a machine-learning approach to automatically generating ontologies from embedding representations.

Latest Patents

Kohama is associated with 1 patent: “Automatic Ontology Generation by Embedding Representations.” The disclosed system receives a vectorized feature set derived from an embedding and processes it through multiple fusers to generate a representation. It then derives tasks from that representation, assigns each task qualifiers such as a weight value, loss function, and feedforward function, and computes weighted losses. By backpropagating those losses through the fusers, feature set, embedding, or a combination of them, the system produces a data model.

Career Highlights

Kohama’s listed company affiliation is Mercari, Inc. The patent describes a method for using embedding representations and task-specific learning objectives to support automatic ontology generation.

Collaborations

Sho Arora and Jeffrey Kenichiro Hara are identified as Kohama’s coworkers. The available information does not specify their roles in the patent or whether they collaborated on it.

Conclusion

Shotaro Kohama’s patent presents an approach to generating data models for ontology creation by combining embedding-derived features, multiple fusers, and weighted task losses.