Interface by Reinforcement Learning
Shun'ichi TANO, Mitsuru TSUKIYAMA
Graduate School of Information Systems,
University of Electro-Communications
1-5-1 Chofugaoka, Chofu-shi, Tokyo 182-8585, JAPAN
Tel:+81.-424-43-5601, Fax: +81-424-43-5681
The interface based on the pen and the paper is so natural that most of the people can easily use it. We can use it without any perceptual load. Therefore the usage of the combination, i.e. the pen and the paper, does not interrupt our thinking process ( Wilcox 1997, Tano 1997). However the current pen-based user interface is far from it. For example, concerning the handwriting, we are usually forced to write in the prefixed area or to write each character separately so that the system can distinguish each character correctly. Concerning the pen gesture, we are usually forced to be familiar with the implemented gesture commands.
Our goal is to realize the natural pen-based user interface that does not disturb the thinking process by adapting itself to the user preference ( Tsukiyama and Tano 1998).
We have extracted two serious problems from many drawbacks, which often cause the interruption of the human's thinking process.
First one is the weakness in grouping of the handwriting strokes. It is often the case that the system fails to group the handwriting strokes correctly and eventually can not recognize it. It is well known that the recognition rate of Japanese characters is quite high if the grouping is completed adequately. If the