||With the widespread of computer software in recent decades, software patent has become controversial for the patent system. Software patents may easily fall into the gray area of abstract ideas, whose allowance may hinder future innovation. However, without a precise definition of abstract ideas, determining the patent claim subject matter eligibility is a challenging task for examiners and applicants. In this research, we address the software patent eligibility issues by proposing an effective model to determine patent claim eligibility and examine the patent examination process to predict patentability. Furthermore, with patent claim features and important prosecution events, we attempt to identify important indicators to valuable patents. |
We collect patent claims, patent examination records, and patent litigation data of software patents from USPTO website, USPTO PAIR, Google Patents, and MaxVal's Patent Litigation Databank. The experiment results show our patent claim eligibility model reaches the accuracy of more than 80%, and domain knowledge features play a crucial role in our prediction model. Using sequence learning on patentability, our patentability predictive model can achieve around 90% accuracy based on our time-duration features. With the value indicators identified by previous models and prior studies, the accuracy of our patent value model can reach up to 88%.