What forms of verbs (potential or passive) are more frequent in Japanese? A human can tell which form is which, can the machine do it? So, the information about which form is more frequent is a good start. Later, the contex sensitive rules can be implemented as well, but everything has its limitations.
network profile
