Abstract. We propose a semantic matching network for the matching of cursive Chinese handwritten annotations. This architecture combines the semantics of Chinese language with the traditional elastic ink matching. Using semantics can make the matching algorithm more intelligent by pre-selecting the most likely candidates before elastic ink matching is applied thus speed up the whole matching process. The semantic matching network can also establish a link between Chinese handwritten annotations and typed text, which can be used to match between these two. Ou, experiments show that 75 -85% recall can be achieved with a speed improvement of 85% over traditional elastic ink matching. IntroductionPen computers and PDA's have been existing for more than a decade. The use of stylus in a pen computing system has already brought advantages in many ways, but handwriting recognition (HWX) proved to be a more difficult problem than most people first expected. Instead of HWX, some research has been done on electronic ink matching [5] that tries to match a query against raw electronic ink data without attempting to recognize them. Similar research can also be found in the works of [10] and [11], but no applications were identified. Pavlidis et al [9] has used shape metamorphosis to recognize on-line handwritten patterns, but their work is limited to handwritten patterns with small number of stroke segments mainly single words or simple shapes. The traditional elastic matching incorporating a dynamic programming procedure was previously described by Tappert [12] on its applications in on-line handwriting recognition. A variation of elastic matching was proposed by Lopresti et al. [4] and it has been used on the matching of handwritten annotations based on stroke level. This algorithm can achieve high accuracy rate in spite of the variations of the way people write. This is specially appropriate for handling informal cursive Chinese handwriting. Yet it was quite slow.While the traditional elastic matching algorithm matches handwriting without attempting to recognize it, a human, on the other hand, might take a different approach. As illustrated in Figure 1, when a human tries to match two pieces of handwriting, he/she first tries to identify their semantics, then matches them. In Chinese writing, radicals, which are small structural parts of a character, are basic elements of semantics. They usually have their own meanings.

No abstract

No abstract