On the Issue of Optimum Machine Learning Methods for Filling and Updating Nuclear Knowledge Graphs

The paper deals with the issues of finding and researching optimum algorithms for classification and semantic annotation of textual network content in the interests of filling and updating nuclear knowledge graphs in Russian and English. Testing of the studied algorithms is carried out by the method of cross-validation. The novelty of the presented research is due to the application of the Pareto's optimality principle for multi-criteria evaluation and ranking of the studied machine learning algorithms, provided that there is no a priori information about the comparative importance of the criteria. The features of the software implementation of efficient classification and semantic annotation algorithms as part of a scalable semantic web portal hosted on a cloud platform are discussed. The proposed software solutions are based on cloud computing using DBaaS and PaaS service models to ensure the scalability of data warehouses and network services.