Research On Association Discovery Based On Data Lineage In Dataspace

With the development of information technology, data information is gradually showing the characteristics of mass, heterogeneity and complexity. The traditional database technologies fail to manage such data in an efficient manner. Now, Dataspace is proposed to solve the problem of heterogeneous data, and it emphasizes the association and evolution between data. Patent literatures contain rich structured and unstructured information.Therefore, analyzing the mass patent data and mining potential association have become a research hot spot.Due to the lack of citation in the patent literatures, the author cited motivation is difficult to determine, as a result, it is difficult to use the citation relationship as an evaluation metric of patent technology association. In order to solve this problem, this paper constructs an integrated semantic similarity model between patents. Firstly, according to the structured information of patents, we build the same author relationship matrix and the same IPC relationship matrix. Secondly, we extract subject terms from patent title, abstract and claims to build the patent text similarity matrix. Finally, the integrated semantic similarity model is built through multi-dimensional integration.Next, we incorporate a temporal factor to the integrated semantic similarity model to build association network of patent lineage. Firstly, by making use of potential citation relationships in association network of patent lineage, the patent value evaluation algorithm is proposed based on the patent value decay over time and the contribution of cited patents.Following that, in order to avoid recalculating the contribution of the new patents to the original patents, an efficient update algorithm of patent value is proposed.Finally, the experiment results show the accuracy of the integrated semantic similarity model and the efficiency of dynamic update algorithm of patent value.