汉字识别误差校正

本文提出了一种对句法模式进行误差校正的方法,通过分析变形笔划y与文法G能接受的笔划x,求出y与G的最小误差校正距离|J|,比较y与其它文法G′的最小误差校正距离|J′|,确定y应属于G或G′。为了改善误差校正剖析程序的执行速度,规定了一个能减少回溯次数的误差校正算子的优先级。同时,用启发式方法进行搜索以减少搜索的深度。最后给出实验结果。

Error-Correcting for Chinese Character Recognition

A method of error-correcting for syntactic pattern recognition is introduced. By analysing the deformated stroke y which is represented in a string and the string x which may be deduced from grammar G, the minimum error-correcting distance between y and G can be gotten. Comparing this with the distance between y and grammar G', we can conclude whether the string y belongs to G or G', though y can neither be deduced from G nor from G'. In order to reduce the time used by errorcorrecting parsing program, the error-correcting operator precedence and heuistic rule have been developed. which can decrease both the backtracking times and search depth. We end this paper with some experimental results.