Mapping Patent Topics and Technological Relatedness in Next-Generation Information Technology: Evidence from the Chengdu-Chongqing Economic Circle
TL;DR
A layered interpretation of the regional portfolio spanning digital infrastructure, equipment, sensing, system integration, and enabling materials is supported, supporting a layered interpretation of the regional portfolio spanning digital infrastructure, equipment, sensing, system integration, and enabling materials.
Abstract
Patent-based evidence is widely used to map regional technological capabilities, but topic identification is often separated from the analysis of technological relatedness. This study examines next-generation information technology in the Chengdu-Chongqing Economic Circle using 90,232 patent records downloaded from the China National Patent Database and covering 2014-2024. Chinese patent texts were encoded with the Sentence-BERT checkpoint all-MiniLM-L6-v2 and analyzed through BERTopic, UMAP, HDBSCAN, class-based TF-IDF (c-TF-IDF), cosine similarity, and hierarchical clustering. The six substantive topics contain 78,511 records; 11,721 observations assigned to the outlier/noise class were excluded from topic-level analysis. The identified topics are communication networks and data architecture (52,707), device structure and installation (10,949), optical sensing and imaging (7,689), metal preparation and processing (2,866), intelligent systems and information integration (2,308), and graphite development and application (1,992). Communication networks account for 67.1% of the classified records and show the strongest pairwise similarity with device technologies (0.96). Optical sensing forms an intermediate hardware-perception branch, while the two material topics are more distinct. Intelligent systems are strongly related to communication and device topics in pairwise comparisons but remain globally differentiated in the hierarchical structure. The findings support a layered interpretation of the regional portfolio spanning digital infrastructure, equipment, sensing, system integration, and enabling materials.