Adaptive Correction of Errors from Segmented Digital Ink Texts in Chinese Based on Context

Abstract

Digital ink texts in Chinese can neither be converted into users’ desired layouts nor be recognized until their characters, lines, and paragraphs are correctly extracted. There are many errors in automatically segmented digital ink texts in Chinese because they are free forms and mixed with other languages, as well as their Chinese characters have small gaps and complex structures. Paragraphs, lines, and characters (recognizable language symbols) in digital ink may be wrongly extracted. An adaptive approach based on context is proposed to correct wrongly extracted these objects. Each extracted object is first adaptively visualized by color and shape labels according to relations between it and its neighbors. Users use simple gestures naturally and easily to merge and split wrongly extracted objects. Contexts are constructed from users’ gestures and objects invoked by them, where users’ intensions are identified. We have conducted experiments using real-life segmented digital ink texts in Chinese and compared the proposed approach with others. Experimental results demonstrate that the proposed approach is feasible, flexible, effective, and robust.