Home

My co-authored article, “Classifying Patents in Spanish within the Subclasses of the Standardized International Patent Classification System: A Case Study”, has just been published. This work was presented at the Tenth International Conference on eDemocracy & eGovernment (ICEDEG 2024) in Lucerne, Switzerland.

Based on the master’s thesis of my former student, Daniel Atiencia, from the Pontificia Universidad Católica del Ecuador, the article offers valuable insights into patent classification. It was a pleasure to host Daniel in Switzerland and provide him with the opportunity to experience his first academic conference.

The article is now accessible in the IEEE Xplore Library alongside other exceptional papers from ICEDEG 2024.

Abstract:

This applied research project investigates the application of data mining and deep learning methods to construct a supervised learning framework tailored for patent classification within the International Patent Classification system. The focus is on Spanish-language patents retrieved from the PATENTSCOPE database. A novel approach employing a Separable Convolutional Neural Network, a streamlined variant of the traditional Convolutional Neural Network, is adopted to minimize the parameter count and computational complexity within the convolutional layers. This approach leverages the inherent sequential nature of word order to extract semantic meaning from individual words within the patent titles and abstracts. To further enhance performance, a pre-trained Word2vec model is incorporated. Word2 vec injects prior linguistic knowledge into the training process, improving the model’s ability to represent and understand the relationships between words. Initial findings demonstrate the model’s ability to provide insightful patent classifications, facilitating analysts and researchers in navigating diverse technological domains while also serving as a valuable resource for identifying potential innovation avenues.