统计与规则相结合的维吾尔语句子边界识别

Sentence boundary detection of Uyghur based on rules and statistics

句子边界识别是词性标注和句法分析等自然语言处理系统的基础问题。提出了一种统计与规则相结合的维吾尔语句子边界识别方法,首先利用歧义段落分类算法分类段落,第二步对无歧义段落进行基于规则的句子边界识别,最后使用最大熵模型对有歧义段落进行句子边界识别。该方法有效利用规则弥补最大熵模型因数据稀疏而误判不存在任何歧义情况的不足,使用最大熵模型有效地消除歧义,提高算法的鲁棒性,召回率达到了98.77%。

Sentence boundary is an important initial task for many natural language processing applications,such as part-of-speech tagging and parsing etc.This paper proposes an automatic sentence boundary detection method of Uyghur based on rules and statistic.Firstly,the paragraph detecting algorithm classifies the ambiguous and unambiguous paragraph.In the second step,the rule based sentence boundary detector process the unambiguous paragraphs.Finally,the maximum entropy based sentence boundary detecting model identifies the ambiguous paragraph sentences.This method improves robustness of the method by making plenty use of rule to reduce the failure of the ME model to identify the unambiguous paragraphs which can be attributed to the sparsity of the training data used and the ME model to resolve ambiguity,the recall of this method reaches 98.77%.