【作者】 王洁；

【导师】 宋柔；

【作者基本信息】 北京语言大学 ， 语言学及应用语言学， 2005， 硕士

【摘要】 汉语多音词自动标音的研究可以为语音合成、汉语教学及音字转换评测等领域服务。本文采用N元文法的统计方法对现代汉语中的多音词进行自动标音。N元文法方法存在的一个突出问题是数据稀疏,而现有的解决数据稀疏的平滑方法并不是很理想。本文利用三元模型,通过引入相似词,采取“词形-相似词-词性”三步回退的策略,比较好地缓解了数据稀疏问题。实验结果说明这个方法是有意义的。本文从四个方面分析了该方法的可行性;并通过两个简单的实验预期了该方法在排歧领域的通用性。基于这样一种方法还可以做很多有意义的工作。更多还原

【Abstract】 The research of automatic phonetic notation on Chinese Polyphonic words can be applied to TTS, Chinese Teaching and Phoneme-to- Grapheme conversion test.The key problem in N-gram method is the problem of sparse data which still can not be solved effectively now. Based on trigram models, this paper proposes a three-step method of "word—similar word—part of speech" by incorporating the similar words and solves the problem of sparse data to a large extent.The result is meaningful. The feasibility of the method can be proved by four aspects. The versatility of the method can be expected in other kinds of work about disambiguation.Based on this method, we are able to do some further interesting work.更多还原