本文从书面语和口语存在的差异出发,提出了语言模型的语体自适应方法。自适应采用了几种不同的计数意义上的插值算法。考虑Katz平滑的插值算法根据trigram单元的可信度来分配权值。基于trigram语体特征分类的自适应算法根据trigram单元的语体特征倾向动态分配权值,并选取了几种不同的权值生成函数。对口语语料做音转字的实验证明,使用这几种自适应算法可以让基准模型的性能有不同程度的提高,其中综合考虑单元可信度和特征倾向的算法效果最好,相对于本文的两个基准的汉字错误率下降率分别达到了50.2%和23.7%。
Abstract
In this paper, a language style based adaptive method for language model is proposed based on the differences between oral and written languages. Several interpolation methods based on trigram counts are used for the adaptation. An interpolation method considering Katz smoothing computes weights according to the confidence score of a trigram. An adaptation method based on the classification of a trigram’s style feature computes weights dynamically according to the trigram’s language style tendency with several weight generation functions proposed. Experiments on spoken Chinese corpora show that these methods could reduce the Chinese character error rate for pinyin-to-character conversion to some extent, more or less, and the one considering both a trigram’s confidence and style tendency achieved the best performance with character error rate reduction of 50.2% and 23.7% , respectively, compared with two baselines in this paper.
关键词
计算机应用 /
中文信息处理 /
统计语言模型 /
trigram /
自适应 /
语体 /
插值算法
{{custom_keyword}} /
Key words
computer application /
Chinese information processing /
statistic language model /
trigram /
adaptation /
language style /
interpolation method
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Gengqing Wu, Fang Zheng. A Method to Build a Super Small but Practically Accurate Language Model for Handheld Devices[J]. J. Computer Science & Technology, 2003, 18 (6) : 747 - 755.
[2] Fang Zheng, Zhanjiang Song, Pascale Fung, et al. Mandarin Pronunciation Modeling Based on CASS Corpus [J]. J. Computer Science & Technology, May 2002, 17 (3) : 249 - 263.
[3] R. Rosenfeld, et al. Error Analysis and Disfluency Modeling in the Switchboard Domain [A]. In: proceedings of the 4th International Conference on Speech and Language Processing (ICSLP) [C]. Philadelphia, PA, USA, 1996.
[4] 吴根清,郑方,金凌,等,一种在线递增式语言模型自适应方法[J]. 中文信息学报, 2002, 16 (1) : 60 - 65.
[5] Rukmini M. Iyer, Mari Ostendorf, Modeling long distance dependence in language: topic mixtures versus dynamic cache models[J]. IEEE Transactions on Speech and Audio Processing, Volume 7 Issue 1, Jan 1999. Page(s) : 30 - 39.
[6] Daniel Gildea, Thomas Hofmann. Topic Based Language Models Using EM[A]. In: proceedings of 6th European Conference on Speech Communication and Technology (Eurospeech′99) [C]. 1999, pages 2167 - 2170.
[7] R. Rosenfeld. A Maximum Entropy Approach to Adaptive Statistical LanguageModel[J]. Computer Speech & Language, 1996, 10: 187 - 228.
[8] S. M. Katz. Estimation of Probabilities from Sparse Data for the Language Model Component of a Speech Recognizer[J]. IEEE Transaction on Acoustic, Speech and Signal Processing, 1987, 35 (3) : 400 - 401.
[9] Genqing WU, Fang ZHENG, Wenhu WU, et al, Improved Katz smoothing for language modeling in speech recognition[A]. International Conference on Spoken Language Processing 2002 [C]. Colorado, USA, Sep.16 - 20, 2002, pp. 925 - 928.
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
PDF(298 KB)
