Databases and semantic networks for computer design of inorganic materials

At present, hundreds of databases (DBs) on substances and material properties have been developed. The prime aim of their operation is information service. Fifteen years ago, we proposed to use an extensive information of DBs not only for information service but also for searching regularities in data and the application of these regularities for prediction of new substances. The semantic networks of special interest were used to search for regularities. Using these deduced regularities, we have predicted thousands of new compounds in ternary, quaternary and more complicated systems and have estimated some of their properties (crystal structure crystal type, melting point, homogeneity region, etc.). Comparison of our predictions with experimental data, obtained later, showed that the average reliability of predicted inorganic compounds exceeds 80%.