This is exactly what we had in mind. The latency of hitting the server for each kanji conversion would make it difficult to compete with a native implementation, but this is just a hack to fill in the holes when no such implementation exists.
Aside from other language support, we're thinking of eventually implementing a passive feedback mechanism that would tell the back end about conversion preferences, so that users could teach the system merely by correcting suggestions.
Comments
This is exactly what we had in mind. The latency of hitting the server for each kanji conversion would make it difficult to compete with a native implementation, but this is just a hack to fill in the holes when no such implementation exists.
Aside from other language support, we're thinking of eventually implementing a passive feedback mechanism that would tell the back end about conversion preferences, so that users could teach the system merely by correcting suggestions.