会話型識別木作成システム

概要

論文の詳細を見る
A man-machine interactive design tool called TBS has been developed, which, with a storage-type display and a keyboard, helps a human designer to build intuitively a decision tree for character recognition. A recognition tree has been formulated by distinguishing node data from procedures called feature-routines.The TBS node modification section works on the node data and allows insertion,deletion,and replacement of nodes in a recognition tree. The TBS recognition diagnosis section either allows to make a performance-test of an obtained tree or provides helpful information many a way of the selected features working on the actual data. In fact, three recognition trees have been built, one of which, KATAKANA tree, has a size of 2,000 nodes and achieved 98% of correct recognition date.The TBS has been evaluated through experiences of building large recognition trees. Some future problems in intuitive design of a recognition tree are also discussed.
一般社団法人情報処理学会の論文
1978-02-15