Tracking the Evolution of Product and Service Inventions with Trademark Data

Abstract

In the “identification” section of trademark documents, assignees need to clearly describe goods and services covered by their trademarks in ways that general readers can easily understand. We use Natural Language Processing (NLP) to identify the first appearance and subsequent follow-ups of product inventions disclosed in trademark identifications, and then propose several measures to evaluate the novelty of these product inventions. Our novelty measures offer timely information on invention value as they are not only available on filing dates, but also positively associated with trademarks’ survival duration. More importantly, we construct a unique identifier that disambiguates all trademark assignees of trademarks, which allows us to evaluate companies’ product inventions based on their trademark portfolios. We plan to make our datasets publicly available to help the research community explore the untapped potential of trademark documents.

Keywords: trademarks; Natural Language Processing; inventions; product innovation; intellectual property.

JEL Classification: O30; O34;