Analytics in Post-Grant Patent Review: Possibilities and Challenges (Preliminary Report)

Recent analysis of litigation outcomes suggests that nearly half of the patents litigated to judgment were held invalid. Commonly available patent search software is predominantly keyword based and takes a “one-size-fits-all” approach leaving much to be desired from a practitioner’s perspective. We discuss opportunities for using text mining and information retrieval in the domain of patent litigation. We focus on post-grant inter partes review process, where a company can challenge the validity of an issued patent in order, for example, to protect its product from being viewed as infringing on the patent in question. We discuss both possibilities and obstacles to assistance with such a challenge using a text analytic solution. A range of issues need to be overcome for semantic search and analytic solutions to be of value, ranging from text normalization, support for semantic and faceted search, to predictive analytics. In this context, we evaluate our novel and top performing semantic search solution. For experiments, we use data from the database USPTO Final Decisions of the Patent Trial and Appeal Board. Our experiments and analysis point to limitations of generic semantic search and text analysis tools. We conclude by presenting some research ideas that might help overcome these deficiencies, such as interactive, semantic search, support for a multistage approach that distinguishes between a divergent and convergent mode of operation and textual entailment.