The recent paper published in PLOS ONE presents a novel approach for efficiently retrieving a substantial number of patents related to specific technologies. The authors enhance an automated patent landscaping algorithm by incorporating a manageable amount of human supervision to enhance result accuracy and consistency. The effectiveness of the approach is demonstrated through its application to six emerging technologies: additive manufacturing, blockchain, computer vision, genome editing, hydrogen storage, and self-driving vehicles.
The research article presents a comprehensive and systematic approach to identifying technology clusters through automated patent landscaping. The integration of human supervision, detailed methodology, and thorough evaluation contribute to the robustness of the proposed method. The external validation adds credibility to the results, and the study holds implications for understanding the landscape of emerging technologies.
Table of Contents
Human-supervised algorithm
The authors introduce the need for a systematic and accurate method to retrieve patents related to specific technologies. They highlight the incorporation of human supervision to refine an automated patent landscaping algorithm, aiming to improve the reliability of the results. The study focuses on six technologies deemed representative of current advancements.
Using an anti-seed for precision
The researchers propose an extended approach to address limitations in existing methods. This includes augmenting the anti-seed with challenging examples derived from human labeling of seed patents. The article delves into the use of transformers for text classification, acknowledging their computational costs. The training and evaluation of models are performed on distinct train-test sets for each technology.
High levels of consistency
The deployment of the algorithm at the patent family level is discussed, offering advantages in terms of coverage and computational tractability. The article provides insights into the algorithm’s performance, consistency, and robustness through detailed evaluation metrics. Notably, the expansion step shows a high level of consistency, reinforcing the relevance of the delimited technology clusters.
The authors use the algorithm’s output to validate results and investigate the sense-making aspect. They leverage the PatCit database to examine the most cited academic papers by technology, providing a link between research articles and patents. The section emphasizes the importance of external validation in ensuring the reliability of the approach.
Encouraging results
The evaluation of the performance and consistency of the extended patent landscaping approach is deemed encouraging. The authors stress the significance of their results in understanding technology clusters and
