Artificial intelligence (AI) and machine learning techniques are changing the world of patent data analysis.
A patent document is a well-structured document, with title, abstract, claims, and description all clearly defined. Patents and applications are also classified using one or more standardized classification codes. These facts make patent data very suitable to be processed with machine learning techniques. Indeed, computer science researchers and service providers in the patent industry have been using AI in patent data analysis for a long time. For example, patent landscapes created with the assistance of text clustering methods were already commonly available before 2008. However, 10 years ago, AI methods didn’t play any important role in patent analysis because a large number of people either didn’t pay enough attention or just didn’t think about it; and for those who did, they often felt that the algorithms were difficult to implement and the results were not very accurate. In short, patent data was ready for AI, but not the other way around.
Over the past few years, not only have AI techniques been greatly improved, but people have also become more aware of them. As a result, using AI to process patent data has become a lot easier and more accurate. It becomes the rational choice of the next step for patent analysis process and product development. For example, a programmer or a patent analyst can have easy access to enterprise-grade computing power and well-documented and easy-to-use programming libraries. Patent search, patent review, and many other type of patent work are being changed by modern AI technology.
At yet2, we started to use machine learning and other statistical methods quite earlier. We built a statistical model to evaluate a patent’s value based on its likelihood of being litigated. By studying attributes of litigated patents, we discovered they look significantly different to other patents in some aspects. We test our model on newly litigated patents, and portfolios acquired by some non-practicing entities (NPEs), and always find the results very convincing. Our patent analysts have been using this model in portfolio prioritization and valuation projects, and our clients have found using such high-quality AI assistance in patent analysis projects provide great added value. We update this model regularly with new data.
We have also used modern AI techniques in creating patent landscape reports, categorizing patent portfolios, filtering patent search results, and providing patent publication monitoring services. AI techniques often provide a quick and relatively accurate way to complete a first pass in a patent analysis project. A human expert can then review and utilize the initial results and continue the analysis at a deeper level. We also plan to introduce AI in identifying interesting patent transactions later this year, which will be another feature of our patent transaction monitoring services.
AI techniques for patent analysis still have a long way to go, but with more resources being put into it, we look forward to leveraging AI and machine learning for patent services in more creative and efficient ways!
