Ontology-Driven Multi-Source Heterogeneous Data Integration Using SSN-SAO Framework for Patent Similarity Analysis

Abstract

Differences in data models and standards make it difficult to directly compare, exchange, or share multi-source, heterogeneous data. Queries between data sources are also challenging, resulting in low data utilization. Therefore, this study proposes a multi-source heterogeneous data integration technology based on the concept of ontology. First, the multi-source heterogeneous data semantic integration technology based on semantic sensor networks is analyzed. Then, based on this, the subject-verb-object semantic structure is introduced and analyzed. The patent representation target is identified by combining this structure with the text characteristics of patent literature. Finally, analyze the results of the proposed technology. The results showed that the recall rate of the integrated technology, which combined semantic sensor networks and subject-verb-object semantic structures, ranged from 64.8% to 68.1%. Its F value was higher than that of the comparison technology, followed by AGDISTIS, and TagMe2 had the lowest. As the number of candidate individuals increased, the precision rate gradually rose, reaching up to 82.6% at its peak. When applied to patent processing, threshold combinations 1, 2, 5, and 9 performed better. Among them, the proportion of patent similarity repetition values in threshold combination 9 was 18%, and the proportion of patents with a similarity of 0 was 46%. After checking the patent content, it was found that its measurement results were the most accurate. Consequently, the proposed technique not only delivers state-of-the-art performance but also markedly elevates the exploitation of multi-source heterogeneous data, furnishing a robust technological backbone for both scholarly inquiry and practical deployment across relevant domains.