关键词: 同类反应, 基核反应, 结构描述符, 化学超空间

Although a huge amount of reaction data are available, chemists still feel inconvenient to find relevant information from these data. The main reason is that the current reaction databases are based on substructure searching which is an approach quite different from that chemists often use. This problem is solved by a novel approach, which acquires reaction knowledge finely described in a layer model by executing twice classification on the reactions. In the first layer classification the reactions are classified in a hyperspace created based on a set of widely applicable reaction descriptors named first order reaction descriptors. On the basis of the results of the first classification, we can extract the common structural characteristics that are used as the reaction descriptors in the second layer classification and lead to a finer classification. Thus, we can obtain the reaction cores from the second finer classification of the reactions. By integration of specific reactions, reaction cores and reaction types, the reaction knowledge can be more rationally organized in a generic reaction knowledge base with better utilization.

Key words: generic reaction, reaction core, structure descriptor, chemical super-space