Prior Art Candidate Search on Base of Semantic Tree Analysis
Artyom Bobunov, Dmitriy M. Korobkin, Sergey Alekseevich Fomenkov · 2025
The article discusses approaches to searching for analogous patents based on key phrase search, as well as by comparing semantic trees representing the component structure of a patented invention. Particular attention is paid to the use of graph databases, such as Neo4j, to represent complex relationships between components of patented devices. Approaches to searching for patent analogs based on key phrases are described. This method allows you to quickly identify relevant documents containing similar concepts in the form of text phrases. At the same time, comparing semantic trees provides a deeper analysis of the component structure of patents, including the relationships between their elements. The advantages of graph structures in the context of semantic information processing are considered, including their flexibility, scalability and the ability to identify hidden patterns. An example of constructing semantic trees and comparing them for two patents is given, illustrating the applicability of the method. As a result, it is shown that the combination of text search based on key phrases and analysis of graph representations of component structures of a patented invention allows to significantly increase the accuracy and efficiency of searching for patent analogues. This opens up new prospects for the development of intelligent data processing systems in the areas of patenting, artificial intelligence and big data analysis.