Logo image
Using un-supervised machine learning approach to generate knowledge ontology for patent analytics
Thesis

Using un-supervised machine learning approach to generate knowledge ontology for patent analytics

Zou, Chen-Han
Masters, 國立清華大學, 工業工程與工程管理學系所
2017

Abstract

專利分群 潛在狄利克雷分配 本體論 patent clustering Latent Dirichlet Allocation ontology
The growth of global patenting activities has been phenomenal in recent decades due to rapid technology development and enterprises seeking protection for their technical innovation. This research aims to develop a novel methodology of "intelligent intellectual property (IP) topic e-discovery." The intelligent IP topic e-discovery will track technology development key topics dynamically and automatically. This system will be built into a computer-supported IP topic e-discovery platform to support R&D planning and IP strategies. This research searches related patents through paid or free patent database, e.g. Derwent Innovation Index platform and USPTO. The first step is to use smart clustering methods to separate patents into key groups. Then, Latent Dirichlet Allocation (LDA), an unsupervised machine learning (M/L) approach, will be investigated (and completed with other methods). The topic models are constructed and their key technical terms under each topic are discovered. Finally, the project will extract important key terms and conduct the process of ontology construction, using the fundamental concept of hierarchical LDA. Related technical and functional terms can be visualized on the ontology maps (i.e. domain knowledge maps) to help enterprises analyze the target technology developing trend in patent portfolios.

Metrics

1 Record Views

Details

Logo image