New AI tool gauges patent potential of scientific papers before commercialization

League of Scholars in Sydney introduced the Translation Readiness Index, an AI-driven tool that evaluates the patent-readiness of scientific publications by comparing their wording to that of earlier patent-associated papers. The system processes only titles and abstracts, feeding them into multiple classifiers; the top performer correctly ranks patent-linked studies 78% of the time. Researchers trained the model on 20,610 papers, including 9,431 with known patent connections, and identified characteristic terms such as “prototype,” “device,” and “design.”

A focused test on the 100 highest-scoring papers from the University of Western Australia revealed that 83 featured industry-linked co-authors and 34 involved authors with prior patents. While co-founder Paul McCarthy cautions against using the index as a sole investment guide, he suggests it could highlight unexpected opportunities. Other scouting tools like Haystack are already being used at institutions such as Cornell University to sift through thousands of papers annually.