DATA-DRIVEN PRODUCT DEVELOPMENT: PATENT DATA ANALYSIS USING NATURAL LANGUAGE PROCESSING

Abstract

The product development lifecycle exhibits many big data flows of internal or external sources and destinations. Until recently, means of analyzing these data flows were severely limited due to performance and storage limits. With the advancement in technology, one can utilize these data flows to improve the product development process thus yielding better results. This thesis finds literature related to big data flows in the product development process and then classifies these flows and their position in the process. It also discusses the challenges and opportunities of utilizing big data analytics in the product development process. This thesis also aims at developing a novelty measure for patents, a specific data flow inside the lifecycle, which is a basis to measure the level of patent innovation. Patents are a main proxy of invention, and each patent exhibits varying inventive value and novelty compared to the corpora. Prior studies of patent novelty have suggested a citation-based approach to measure the novelty across patents. Despite their progress in measuring patent novelty, several challenges remain: The inability to consider single class inventions, and the inclusion of patent-only citations. To address these challenges, we devise a novel approach using NLP techniques to find a text-based novelty measure. The proposed method is applied on patents that belong to a common category, which represents a subset of patents under a specific patent class. We then extract the novelty-value profile of those patents and discuss a use case for product development – extracting patent novelty and predicting inventive value. Product developers would benefit from our proposed approach by allowing them to predict value of patents being developed. These findings would contribute to having an alternative way to measure novelty, which complements previous citation-based methods. Future research would build upon text-based measures which will further improve data-driven approaches tackling novelty assessment.