An evaluation of online databases by methods of lattice theory

The evaluation of databases is of great importance for potential users. The usual technique to define a ranking index however is not helpfull because users may have different interests concerning databases. Therefore evaluation techniques are needed, which do not mask different elements of information performing a ranking. The appropriate technique is given by the theory of order relation which leads to partially ordered sets. Partially ordered sets can be visualized by Hasse diagrams. In this contribution 15 databases are characterised by five quality criteria and the Hasse diagram shows that there are four databases which are best suitable with respect to different aspects, namely CAS, ECDIN, BEIISTEIN and HSDB. These four databases are further investigated by the formal concept analysis. A comprehensive diagram shows the specialities of each of these four databases and logical rules are automatically derived, which show that there are interdependencies between qualities within the context of the four priority databases. For example a good quality with respect to the use of chemicals implies a good quality of the general extent of the database with respect to the chemicals.