节点文献
带拼音纠错的汉语音字转换技术
Chinese pinyin to text translation technique with error correction used for continuous speech recognition
【摘要】 提出了一种基于统计和规则的混合方法来实现汉语音字转换。利用汉语的语法规则,在统计语言模型中采用了两种基于词和词性的混合语言模型。在实验中,将这两种混合语言模型与基于词的语言模型进行了比较。实验证明,在语言模型中引入词性后,提高了音字转换正确率。考虑了出现拼音错误时的音字转换问题,提出了一种拼音纠错方法来纠正错误。实验证明,当拼音正确率高于85%时,这种带纠错的音字转换方法可以提高音字转换正确率。
【Abstract】 This paper makes use of a hybrid statistical and rule approach to realize Chinese pinyin to text translation. With the help of Chinese grammar this paper raises two hybrid statistical language models based on word and parts of speech, and the experiments prove that by this model the accurate rate is improved. This paper puts forward an approach to correct some pinyin errors in the case of pinyin errors. When the accurate rate of pinyin is more than 85%, this approach gives a very satisfactory result in the experiments.
【Key words】 continuous speech recognition; statistical language models; natural language understanding ;
- 【文献出处】 清华大学学报(自然科学版) ,JOURNAL OF TSINGHUA UNIVERSITY(SCIENCE AND TECHNOLOGY) , 编辑部邮箱 ,1997年10期
- 【分类号】TN912.34
- 【被引频次】13
- 【下载频次】236