PatenteGEN: uma solução baseada em aprendizagem profunda para geração de patentes

Abstract

The advance of technology, the solutions and tools inventions, which directly affect the way society lives and has been growing and developing, is linked to the human being’s capacity to innovate and use creativity to solve problems. The protection of intellectual property through the application of patents, not only guarantee the rights of inventors, but helps structuring the expression of ideas and the knowledge application . This work approach the development of a tool, to provoke insights and assist in the patents creation, through generative models, with the OpenAI GPT-2 pre-trained model, using the public patent base of United States Patents Trademark Office(USPTO) as a data source. The work is structured in four steps: data collection, model training, creation of the web application and analysis of the generated results