Chinese characters that are similar in their pronunciations or in their internal structures are useful for computer-assisted language learning and for psycholinguistic studies. Although it is possible for us to employ image-based methods to identify visually similar characters, the resulting computational costs can be very high. We propose methods for identifying visually similar Chinese characters by adopting and extending the basic concepts of a proven Chinese input method--Cangjie. We present the methods, illustrate how they work, and discuss their weakness in this paper.
HLT-Short '08 Proceedings of the 46th Annual Meeting of the Association for Computational Linguistics on Human Language Technologies: Short Papers,93-96