| 基于Trigger的长距离语言模型 |
| |
| 引用本文： | 苏韬,孙甲松,王作英.基于Trigger的长距离语言模型[J].计算机工程与应用,2002,38(18):59-61. |
| |
| 作者姓名： | 苏韬 孙甲松 王作英 |
| |
| 作者单位： | 清华大学电子工程系语音处理实验室,北京,100084 |
| |
| 基金项目： | 国家863高技术研究发展计划项目(编号:863-306-ZD03-02-1)，985重大项目(编号:985校-22-攻关-06) |
| |
| 摘 要： | 文章利用Trigger模型描述长距离的词与词之间的关系,并将其与目前通用的N元文法语言模型相结合,同时通过引入参数———置信度来提高Trigger模型得分的可靠性,在汉语自动语音识别系统中实现了一个可实用化的长距离语言模型。实验表明,二者的结合在一定程度上克服了N元文法语言模型描述距离小于N的缺点,提高了系统的识别率。
|
| 关 键 词： | Trigger对 置信度 长距离语言模型 |
| 文章编号： | 1002-8331-(2002)18-0059-03 |
| 修稿时间： | 2002年6月1日 |
A Long-distance Language Model Based on Trigger |
| |
| Su Tao Sun Jiasong Wang Zuoying.A Long-distance Language Model Based on Trigger[J].Computer Engineering and Applications,2002,38(18):59-61. |
| |
| Authors: | Su Tao Sun Jiasong Wang Zuoying |
| |
| Abstract: | In this paper,a Trigger model is exploited to capture the long distance dependence between words.By inte-grating a trigger model with a general N-gram language model and introducing a parameter-confidence to enhance the reliability of score of the trigger model,a practical long-distance language model is implemented for speech recognition.Experiments show that the cooperation of these two models can overcome the shortcoming of N-gram model that it only can describe the word pairs being less than N words apart,and it also improves the word right rate of the system. |
| |
| Keywords: | Trigger pairs confidence long-distance language model |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
