Abstract
The data for automatic piano is expressed thousands of notes, so we take a long time to edit parameters of each note. Then, we developed efficient support system based on an idea, ″If musical note sequence is similar, performance expression of each phrase is almost the same″. The system consists of two systems. One is ″Searching system″. This system searches similar phrases by using dynamic programming matching. Another is ″Inferring system″. This system infers performance expression of non-editing phrases by using input phrases' data, but the data contains some influence of musical symbols such as dynamic marks. Then, this system tries to remove the influence by referring to databases.