Focus

>125 million patents from all around the world. Bibliographic data from 98 countries, and full text from 38.

We built an AI search tool that works through so-called virtual adversarial training. It is shown positive examples of what type of patents the user is interested in and negative examples of what the user is not interested in. Based on the patents' features that are related to its topic, things like NLP on the text but also a proprietary context analysis model we have developed, the model learns what type of patent the user is interested in from the examples. The search process with our tool can be iterative by giving some examples, retraining the model, seeing how well it works, giving more examples, retraining, etc. or the user can bulk upload lists and train the model in almost no time and get to work.

NLP

Deep learning

Scoring data

Landscaping capabilities

Customized score for every patent family.

Excel and CSV

All fields can be exported and there are no limits other than a cap on the information you can extract per batch so as to not melt the infrastructure.