||A Japanese manufacturer of automotive components
|Problem & Context
||The client IP team needed help in the development of an AI/ML tool to auto-screen and categorize patents in a rapidly emerging technology in autonomous vehicles. The volume of patent data was growing rapidly and a productivity enhancing tool was critical to ensure timely assessment of competitor activity. The client required technology categorization was difficult to automate because there was a significant overlap between categories.
|SPA Approach & Deliverables
- A 3-member SPA team was involved in the development of the AI/ML tool.
- The SPA team received about 20 sample patents from the client along with tags specifying whether the patents are relevant or not, and if they are relevant what technology category they belong to.
- To train the AI/ML tool, SPA team expanded the 20 sample patent data-set to more than 5,000 patents. This data-set formed the training data-set for the AI/ML tool.
- The AI/ML tool was integrated into a globally accessible portal which was then tested by the client with an unseen data-set. The model achieved more than 80% accuracy.
- A mechanical engineer and an AI/ML-algorithm developer