基于N-gram的哈萨克语文本校对系统的设计与实现

摘要

在啥萨克语文本非词查错方面,归纳和总结查错方法,在一定规模的哈萨克语词库的支持下,利用哈萨克语的特点,用哈萨克语词干切分程序和哈萨克语的音节规则,从文本中找出非词错误,再用最小编辑距离算法提供最有可能的候选词.在哈萨克语文本真词查错部分,根据上下文信息,采用基于N-gram的语言模型,利用文本的局部连接同现概率三元语法模型来进行真词查错,再用基于编辑距离的模式匹配方法对真词错误提供纠错建议.实验结果表明,系统的查错与纠错效率较好,实验方案是可行的.%For the section of non-word errors checking in Kazakh text, on the basis of summarising and concluding the errors checking methods and supported by a certain size Kazakh lexicon, in the article we use the characteristics of Kazakh and the stem segmentation program and syllable rules of Kazakh language to find the non-word errors from the text, and then provide the most possible candidate word with minimum edit distance algorithm. In the section of real-word error checking in Kazakh text, according to context information and adopting N-gram based language model, we carry out real-word error checking by using ternary grammar model of local connection co-occurrence probability of the text, and then use the edit distance-based pattern matching method to provide error-correction suggestions to the errors of real words. Experimental results show that efficiency of error checking and error correction of this system is fairly good, the experiment scheme is feasible.