摘要
普通话大词汇量连续语音识别的研究已经进行了十多年,虽已取得了显著进展,但距离广泛应用还有相当的距离。语音识别后处理是将前处理所得到的音节流转换为汉字流的过程。研究发现,语音识别系统的后处理对提高系统性能具有十分重要的意义。人类听觉实验表明,人只能听清连续语音流中70%的音节,剩余的30%是靠上下文知识来猜测理解的。因此,语音识别后处理受到了广泛的关注,得到了越来越深入的研究。
本文主要对普通话大词汇量连续语音识别后处理中的语言模型自适应、解码策略、错误处理等问题进行了研究,主要工作与创新包括以下几个方面:
1.汉语混淆网络算法
首先研究了最小贝叶斯风险解码准则以及基于最小贝叶斯解码准则进行最小字错误率解码的若干方法,例如:基于N-best lists的方法、基于word lattice的方法等。在此基础上,考虑到汉语语言的特点,提出一种构造汉语词混淆网络的算法,对于汉语词格(wordlattice)中的长弧,在强制对齐时根据其发音特点快速有效地加入null弧。实验表明改进的构造汉语词混淆网络进行解码的方法与MAP(Maximum a posterior)解码、先前的各种错误率最小化算法相比,有效地降低了普通话大词汇量连续语音识别词错误率。
汉语中一个词一般由1—4个汉字组成,由不同数目汉字组成的词的发音时间长短差别比较大,造成构造的汉语词混淆网络中包含了大量的null弧。本文提出一种构造汉字混淆网络来获取具有最小字错误率的识别结果假设的方法,这种算法显著地减少了构造的汉字混淆网络中的null弧的数目。实验结果表明这种构造汉字混淆网络进行解码的方法有效降低了识别结果的字错误率。
2.解码结果的错误检测与纠正研究
在普通话大词汇量连续语音识别中,识别结果出错的现象和原因非常复杂。本文首先分析了一些常见的普通话大词汇量连续语音识别结果中的错误及其出现的原因。在此基础上,采用基于转换的学习方法从混淆网络中学习纠错规则,实验表明应用这些纠错规则能够有效降低识别结果的词错误率。
考虑到汉语语言的复杂性以及用于错误纠正规则学习的训练语料集有限,不能覆盖所有的错误现象,本文使用统计的方法进行错误的检测与纠正。具体地,本文提出一种基于支撑向量机SVM(SupportVector Machines)进行错误检测与纠正的框架,首先使用SVM对识别结果假设字串中的每个字进行分类,判断其正确性;接下来对于分类为错误的字基于汉语字混淆网络构造候选字序列,对候选字序列重新打分,选择最高得分的字串作为错误纠正的结果。实验结果表明这种方法能够有效地检测出识别结果中的错误并进行纠正,降低了字错误率。
3.语音识别中的区分性语言模型研究
语言模型自适应是根据不断变化的应用环境,调整语言模型中各种现象出现的概率,以适应不同应用环境的特征。本文将Boosting、Perceptron以及最小化样本风险三种算法用于训练语音识别系统中的N-Gram语言模型,使其对特定领域具有自适应能力。实验结果表明使用这三种算法训练的N-Gram语言模型降低了特定领域的语音识别结果的词错误率。其中Perceptron算法训练的N-Gram语言模型的领域自适应能力最好。所以本文在通用领域的语音识别中,将输入的语音与识别输出的汉语词混淆网络作为训练样本,使用Perceptron算法训练区分性语言模型,并用这种语言模型对汉语词混淆网络重新打分。实验结果表明这种方法有效地降低了识别结果的词错误率。
Mandarin Large vocabulary Chinese continuous speech recognition has been researched for more than ten years. Although there are some achievements in continuous speech recognition research, the distance from widespread application is still very long. The post-processing of speech recognition is a processing which converts Pinyin to Chinese characters. The research shows that the post-processing of speech recognition has very important significance to improve the system performance. The experiment on hearing indicates that human can only hear 70% syllables in continuous speech and understand the remaining 30% using context knowledge. Therefore, the post-processing technique of speech recognition is paid a great attention and conducted in-depth studies.
In this thesis, we will make a deep research on the post-processing technique of Mandarin large vocabulary continuous speech recognition, including language model adaptation, decoding strategy and error handling. The main contributions and innovations are described in details as follows:
1. Chinese confusion network algorithms
At the beginning, we study the minimum bayes risk decoding rule and some minimum word error rate decoding methods. According to the characteristic of Chinese linguistics, we proposed an improved algorithm of constructing Chinese word confusion network. The improved algorithm fleetly adds null arc in confusion set when the long arc with a Chinese character string is forcibly aligned in the process of constructing a confusion network. This improved algorithm were evaluated on 2005 HTRDP (863) Evaluation task, where improved word accuracy performance was observed.
In general, a Chinese word consists of 1-4 single Chinese characters, so its pronunciation time length change quite large. So we proposed an novel Chinese character confusion network algorithm for the purpose of decreasing the number of null arc. Experimental result proves that this algorithm cut the character error rate of recognition results effectively.
2. Research on detection and correction of decoding results
On the basis on analysis of decoding errors and reason, we proposed a method that we use transformation-based learning for learning error correction rules from Chinese word confusion network. Experimental result shows significant improvements over recognition results.
Considering the complexity of Chinese and the limited corpus for learning error correction rules, we use statistical methods to detect and correct decoding errors. In details, we use SVM to classify the decoding results, detect the errors; then we use Chinese character confusion network to correct errors. Experimental result shows that this method can effectively detect and correct decoding errors, and reduce the character error rate.
3. Study on discriminative language model of speech recognition Firstly, we study three discriminative methods of language modeladaptation, including the boosting algorithm, the perceptron algorithm and the minimum sample risk algorithm, and present comparative experimental results on the performance of using different approaches to train discriminative language model on the task of speech recognition. Then we use the perceptron algorithm with best discriminative performance to train discriminative language model for general domain Mandarin large vocabulary continuous speech recognition, and rescore the Chinese word confusion network. Experimental result shows that this method can effectively reduce the word error rate.
引文
[1]L.Rabiner,B.Juang.Fundamentals of Speech Recognition Englewood Cliff,New Jersey:Prentice-Hall,1993.
[2]王炳锡,屈丹,彭煊.实用语音识别基础.北京:国防工业出版社.2005,1-2.
[3]王坚语音识别中的说话人自适应研究.北京邮电大学博士论文.2007.
[4]Kai-Fu Lee,Hsiao-Wuen Hon,Mei-Yuh Hwang,et al.The Sphnix Speech Recognition System.IEEE Proc.of ICASSP' 89.1989,44-445.
[5]Liu-span Lee.A Real-Time Mandarin Dictation Machine for Chinese Language with Unlimited texts and Very Large Vocabulary.IEEE Proc.of ICASSP' 90.1990:65-68.
[6]L.R.Bahl,S.Balakrishnan-Aiyer,J.Bellegarda,et al.Performance of the IBM Large Vocabulary Continuous Speech Recognition System on the ARPA Wall Street Journal Task.IEEE Proc.Of ICASSP'95,Detroit,MI.1995,5:4-41.
[7]P.C.Woodland,C.J.Leggeter,J.J.Odell,et al.The 1994 HTK Large Vocabulary Speech Recognition System.IEEE Proc,of ICASSP'95.Detroit,MI.1995,5:7-73.
[8]P.C.Woodland.The development of the HTK Broadcast News transcription system:An overview.Speech Communication,2002,37:47-67.
[9]颜龙.大词汇量汉语连续语音识别系统中若干问题的研究.北京邮电大学博士论文.2005.
[10]赵以宝.汉语语音识别算法研究.哈尔滨工业大学大学博士论文.1999.
[11]H.Hermansky.Perceptual Linear Predictive(PLP) Analysis for Speech.J.Acoust.Soc.AM.1990,1738-1752.
[12]刘加.汉语大词汇量连续语音识别系统研究进展,电子学报,2000,28(1):85-91.
[13]赵力.语音信号处理.北京:机械工业出版社.2003:20-80.
[14]Harvey F.Silverman,David P.Morgan.The application of dynamic programming to connected speech recognition,IEEE ASSP Magazine,pp.6-25,July 1990.
[15]Hermann Ney,Stefan Ortmanns.Dynamic programming search for continuous speech recognition,IEEE Signal Processing Magazine,pp.64-83, Sep.1999.
[16]Hermann Ney,Stefan Ortmanns.Progress in dynamic programming search for LVCSR,Workshop on Automatic Speech Recognition and Understanding,pp.287-294,1997.
[17]Gray R M.Vector Quantization.IEEE ASSP Mag.1984,1(2) 4-29.
[18]D.H.Klat.Review of the ARPA Soc.Amer.1977,62:1345-1366.
[19]D.R.Paul.Speech Recognition Using Hidden Markov Models.The Lincoln Laboratory Journal.1990,3(1):41-6.
[20]L.R.Rabiner et al.A Tutorial on Hidden Markov Models and Its Applications to Speech Recognition.IEEE Proc,1989,72(2):257-28.
[21]J.Picone.Continuous Speech Recognition Using Hidden Markov Models IEEE ASSP Magazine,1990,7:26-4.
[22]Sadaoki Furui.50 Years of Progress in Speech and Speaker Recognition Research,Ecti Transactions on Computer And Information Technology vol.1,no.2 November 2005,64-74.
[23]B.-H.Juang,L.R.Rabiner.Automatic speech recognition - A brief history of the technology development,K.Brown(Ed.) Encyclopedia of Language and Linguistics,Elsevier,2005.
[24]何湘智.语音识别的研究与发展.计算机与现代化.2002,3:3-6.
[25]张雄伟,陈亮,杨吉斌.现代语音处理技术及应用.机械工业出版社.2003.9-35.
[26]胡敬东.语音识别后处理系统的研究.华南理工大学硕士论文.2000.
[27]Jelinek F.Continuous speech recognition by statistical methods,In Proceedings of the IEEE,1976,64(4):532-556.
[28]S.Young.Statistical Modeling in Continuous Speech Recognition(CSR).Engineering Department,Cambridge University.
[29]A.J.Robinson,G..D.Cook,etc.Connectionist speech recognition of Broadcast News.Speech Communication,2002,37:27-45.
[30]Beverlein,X.Aubert,etc.Large vocabulary continuous speech recognition of Broadcast News.Speech Communication,2002,37:109-131.
[31]G.J.Chen,R.A.Gopinath.A Continuous Speaker-Independent PuTongHua (?)tation System,Proc.International Conference Signal Processing,Oct.1996,321-824.
[32]Liu-span Lee.A Real-Time Mandarin Dictation Machine for Chinese Language with Unlimited texts and Very Large Vocabulary.IEEE Proc.Of ICASSP' 90.1990,65-68.
[33]吕萍.广播新闻语料自动识别系统.中国科学院声学研究所博士后学位论文.2005.
[34]雷建军.噪声鲁棒语音识别中若干问题的研究.北京邮电大学博士论文.2007.
[35]冯哲.大词汇量汉语连续语音的快速识别算法研究.复旦大学硕士学位论文.2000.
[36]赵庆卫,王作英.汉语连续语音识别中上下文相关的识别单元(三音子)的研究.电子学报,1999,27(6):79-82.
[37]Young S.Large vocabulary continuous speech recognition:a Review,IEEE Signal Processing Magazine,1996,13(5):45-57.
[38]W.Chou.Pattern Recognition in Speech and Language Processing.CRC Press.Washington,D.C.,New York.2003.
[39]Young S.,Evermann G.,Hain T.et al.The HTK Book(for HTK Version 3.2.1),Cambridge University Engineering Department,2002.
[40]谢锦辉.隐Markov模型(HMM)及其在语音处理中的应用.华中理工大学出版社,1995.
[41]王之禹.语音识别及其相关研究.中国科学院声学研究所博士后学位论文.2002.
[42]刘秉权.汉语语言模型研究及其在音字转换中的应用.哈尔滨工业大学博士学位论文.2003.
[43]陈浪舟.面向语音识别的高性能统计语言模型的研究.中国科学院声学研究所博士学位论文.1999.
[44]徐志明.统计语言模型研究及其在汉字识别中的应用.哈尔滨工业大学博士学位论文.2001.
[45]赵岩.基于统计语言模型的汉语词法分析研究.哈尔滨工业大学博士学位论文.2005
[46]P.R.Clarkson,R.Rosenfeld.Statistical Language Modeling Using the CMU-Cambridge Toolkit.In.Proc.ESCA Eurospeech-1997.1997.
[47]Daniel JurafSky,James H M.自然语言处理综述.电子工业出版社.2005,15-144.
[48]Christopher D.Manning.统计自然语言处理基础.北京:电子工业出版社,2005,228-230.
[49](?)建毅.基于统计与规则相结合的口语语言理解研究.北京邮电大学大学博士学位论文.2005.
[50]Huang Xuedong,Acero Alex,Hon Hsiao-Wuen et al.Spoken language processing:a guide to theory,algorithm and system development,Prentice HALL PTR,2001.
[51]X.L.Albert.An Overview of Decoding Techniques for Large Vocabulary Continuous Speech Recognition.Computer Speech and Language.2002(16):39-114.
[52]刘刚.基于HMM的文字识别和语音识别中若干问题的研究.北京邮电大学博士学位论文.2002.
[53]O.-W.Kwon and J.Park,Korean large vocabulary continuous speech recognition with morpheme-based recognition units,Speech Commun.,vol.39,no.3-4,pp.287-300,Feb.2003.
[54]张东滨.语音识别的大词表解码策略与语速鲁棒性研究.中国科学院声学研究所博士后学位论文.2005.
[55]谢凌云.实时语音识别系统的快速算法研究.中国科学院声学研究所博士学位论文.2004.
[56]Ravishankar,M.K.Efficient Algorithm for Speech Recognition,Ph.D.Thesis,Carnegie Mellon University,1996.
[57]Ortmanns S.,Eiden A.Ney H.Improved Lexical Tree Search for Large Vocabulary Speech Recognition,Acoustics,Speech and Signal Processing,ICASSP '98 Proceedings,pp.817-820 vol.2,May 1998.
[58]Demuynck,K.,Duchateau,J.,Van Compernolle,D.,et al.An Efficient Search Space Representation for Large Vocabulary Continuous Speech Recognition,Speech Communication,Volume 30,Number 1,pp.37-53,January 2000.
Lidia Mangu.Finding Consensus in Speech Recognition,Ph.D.Thesis,Johns Hopkins University,2000.
[60]H.Nanjo,A.Lee,T.Kawahara.Automatic diagnosis of recognition errors in large vocabulary continuous speech recognition systems.In ICSLP,volume 2,pages 1027-1030,2000.
[61]J.Hirschberg,D.Litman,M.Swerts.Prosodic and other cues to speech recognition failures.Speech Communication,(43):155-175,2004.
[62]J.Hirschberg,D.Litman,M.Swerts.Predicting automatic speech recognition performance using prosodic cues,1st Conference of the North American Chapter of the Association for Computational Linguistics,NAACL 2000.
[63]K.Kukich.Techniques for automatically correcting words in text.ACM Computing Surveys,24(4):377-439,1992.
[64]G.Gorrell,I.Lewin,M.Rayner.Adding intelligent help to mixed-initiative spoken dialogue systems.ICSLP-2002,2002.
[65]I.Bazzi,J.R.Glass.Modeling out-of vocabulary words for robust speech recognition,pages 401-404.ICSLP-2000,2000.
[66]V.Siivola,T.Hirsi(a丨¨)aki,M.Creuts,M.Kurimo.Unlimited Vocabulary Speech Recognition Based on Morphs Discovered in an Unsupervised Manner pages 2293-2296,Geneva,2003.EUROSPEECH 2003.
[67]付跃文,杜利民.语音识别错误的分类分析.计算机应用,2005,25(2):28-32.
[68]Kimberly Voll.A Methodology of Error Detection:Improving Speech Recognition in Radiology.PhD thesis,Simon Fraser University,School of Computing Science,University Drive,Burnaby,BC,Canada,June 2006.
[69]M.Turunen,J.Hakulinen.Agent-based error handling in spoken dialogue systems.In Proceedings of Eurospeech,pages 2189-2192,2001.
[70]Kimberly Voll,Stella Atkins,Bruce Forster.Improving the utility of speech recognition through error detection.In SCAR Annual Meeting,2006.
[71]Arup Sarma,David Palmer.Context-based speech recognition error detection and correction.In Proceedings of the HLT-NAACL 2004,pages 85-88,2004.
[72]M.Jeong,B.Kim,G.Lee.Using higher-level linguistic knowledge for speech recognition error correction in a spoken Q/A dialog.In Proceedings of the HLT-NAACL special workshop on Higher-Level Linguistic Information for Speech Processing,pages 48-55,2004.
[73]Diana Inkpen,Alain D'esilets.Semantic similarity for detecting recognition errors in automatic speech transcripts.In Proceedings of EMNLP,pages 49-56,Vancouver,British Columbia,Canada,October 2005.Association for Computational Linguistics.
[74]Stephen Cox,Srinandan Dasmahapatra.High-level approaches to confidence measure estimation in speech recognition.IEEE Transactions on Speech and Audio Processing,10(7):460-471,Oct 2002.
[75]Stephen Cox,Srinandan Dasmahapatra.A semantically-based confidence measure for speech recognition.In Proceedings of Int.Conf.on Spoken Language Processing,volume 4,pages 206-209,Beijing,China,2000.
[76]Andrew Sears,Clare-Marie Karat,Kwesi Oseitutu,et al.2001.Productivity,satisfaction,and interaction strategies of individuals with spinal cord injuries and traditional users interacting with speech recognition software.Universal Access in the Information Society,1(1):4-15,June.
[77]Lin L.Error-Responsive Feedback Mechanisms for Speech Recognizers.Ph.D.thesis,School of Computer Science,CMU,April,1997.
[78]Christine Pao,Philipp Schmid,James Glass.Confidence scoring for speech understanding systems.In Proceedings of ICSLP,1998,815-818.
[79]Rub'en San-Segundo,Bryan Pellom,Kadri Hacioglu,et al.Confidence measures for spoken dialogue systems.In Proceedings of ICASSP,volume 1,2001,pages 393-396.
[80]Jinjuan Feng,Andrew Sears,Clare-Marie Karat.A longitudinal investigation of hands-free speech based navigation during dictation.Technicalreport,UMBC.2003.
[81]Ruhi Sarikaya,Yuqing Gao,Michael Picheny.Word level confidence measurement using semantic features.In Proceedings of ICASSP,volume 1,2003,pages 604-607.
[82]Eric Brill,Radu Florian,John C.Henderson,,et al.1998.Beyond n-grams:Can linguistic sophistication improve language modeling? In Proceedings of COLING/ACL,pages 186-190.
[83]Paul Carpenter,Chun Jin,Daniel Wilson,et al.Is this conversation on track?In Proceedings of Eurospeech,2001.pages 2121-2124.
[84]Manny Rayner,David Carter,Vassilios Digalakis,et al.Combining knowledge sources to reorder n-best speech hypothesis lists.In Proceedings of the ARPA Workshop on Human Language Technology,1994,pages 212-217.
[85]Yongmei Shi,Lina Zhou.Error Detection Using Linguistic Features.HLT/EMNLP 2005.
[86]Lin.Chase.Error-Responsive Feedback Mechanisms for Speech Recognizers.Dissertation for the degree of Doctor of Philosophy.Carnegie Melon University:1997.
[87]Y.He,S.Young.A Data-Driven Spoken Language Understanding System,Proc.ASRU,2003.
[88]赵力.汉语连续语音识别中语音处理和语言处理统合方法的研究.声学学报,2001,26(1):73-781.
[89]苗传江.HNC句类知识研究.中国科学院声学研究所博士学位论文,2001.
[90]Lidia Mangu and Mukund Padmanabhan.Error corrective mechanisms for speech recognition.In Proceedings of ICASSP,volume 1,2001,pages 29-32.
[91]D.Hillard,M.Ostendorf.Compensating for word posterior estimation bias in confusion networks,in International Conference on Acoustics,Speech,and Signal Processing,2006.
[92]Zhou,Zhengyu / Meng,Helen M./ Lo,Wai Kit(2006).A multi-pass error detection and correction framework for Mandarin LVCSR,In INTERSPEECH-2006,paper 1947-Wed1CaP.12.
[93]Stolcke A,Konig S,Weintraub M.Explicit Word Error Minimization in N-best ListRescoring.Proceedings of EUROSPEECH.1997:163-166.
[94]Lidia Mangu,Brill E,Stolcke A.Finding Consensus in Speech Recognition:Word Error Minization and Other Application of Confusion Network Computer Speech and Language,2000,14(4):373-400.
[95]Xue J,Zhao Y X.Improved Confusion Network Algorithm and Shortest Path Search from Word Lattice.ICASSP 2005.2005:853-856.
[96]V.Goel,Kumar S,Byrne W.Segmental Minimum Bayes-risk Decoding for Automatic Speech Recognition.IEEE Trans on Speech and Audio Processing,2004,12(3):234-249.
[97]Soong F.K.,Lo W.K.Nakamura,S.Generalized Word Posterior Probability (GWPP) for Measuring Reliability of Reconized Words,Proc.SWIM 2004.
[98]Wai Kit LO,Fvnnk K.SOONG,Sntoshi NAKAMURA.Generalized Posterior Probability for Minimizing Verification Errors at Subword Word and Sentence Levels.ISCSLP 2004:13-16.
[99]V.Goel,W.Byrne.Task dependent loss functions in speech recognition:Application to Named Entity extraction.ESCAETRW Workshop on Accessing Information from Spoken Audio,pp.49-53,Cambridge,UK,1999.
[100]余涛.汉语语音识别文本的自动纠错研究..上海交通大学硕士学位论文.2003.
[101]Solen Quiniou,Eric Anquetil.Use of a Confusion Network to Detect and Correct Errors in an On-Line Handwritten Sentence Recognition System.Ninth International Conference on Document Analysis and Recognition(ICDAR 2007).
[102]E.Brill.Transformation-based Error-driven Learning and Natural Language: A Case Study in Part of Speech Tagging.In Computational Linguistics,2 1(4):543-565.
[103]L.Mangu,E.Brill.Automatic Rule Acquisition for Spelling Correction.In Proc.ICML,Nashville,1997.
[104]Lager,Torbj(o丨¨)rn(1999) The μ-TBL System:Logic Programming Tools for Transformation-Based Learning.In Proceedings of the Third International Workshop on Computational Natural Language Learning(CoNLL'99),Bergen,1999.
[105]Vapnik.Nature of Statistical Learning Theory.John Wiley and Sons,Inc.,New York,in preparation.
[106]Filip.Mulier.Vapnik-Chervonenkis(VC).Learning Theory and Its Applications IEEE Trans on Neural Networks.Vol.10,No.5,Sep 1999.
[107]Edgar Osuna,et al.Training Support Vector Machines:an Application to Face Detection.
[108]Chris J.C.Burges.Simplified Support Decision Rules.
[109]Chris J.C.Burges,Beruhard Scholkopf..Improving the Accuracy and Speed of Support Vector Machines.Neural Information Processing Systems,Vol.9,M.Mozer,M.Jordan,& T.Petsche,eds.MIT Press,Cambridge,MA,1997.
[110]P.S.Bradley,O.L.Mangasarian.Feature Selection via Concave Minimization and Support Vector Machines.
[111]Z.Zhou,H.Meng.A two-level schema for detecting recognition errors,Proc.ICSLP 2004
[112]Z.Zhou,H.Meng.Error identification for large vocabulary speech recognition,Proc.ISCSLP 2004
[113]F.Soong,W.K.Lo,S.Nakamura.Optimal acoustic and language model weights for minimizing word verifi- cation error,in Proc.ICSLP,2004,pp.441-444.
[114]宁振江.基于数据增强的汉语词表和语言模型自适应技术.中国科学院声学研究所博士学位论文.2005.
[115]Woosung Kim.Language Model Adaptation for Automatic Speech Recognition and Statistical Machine Translation.Ph.D Thesis.The Johns Hopkins University.2004.
[116]P.Clarkson.Adaptation of Statistical Language Models for Automatic Speech Recognition.PhD thesis,Cambridge University Engineering Department,1999.
[117]J.R.Bellegarda.Statistical Language Model Adaptation:Review and Perspectives,Speech Communication 42,2004.
[118]R.Rosenfeld.Two Decades of Statistical Language Modeling:Where Do We Go from Here,Proc.IEEE,88(8),2000.
[119]R.Iyer,M.Ostendorf.Modelling Long Distance Dependence in Language:Topic Mixtures vs.Dynamic Cache Models.Proc.ICSLP 1996,Volume 1 pp.236-239
[120]Brugnara,Fabio/Federico,Marcello(1996):Techniques for approximating a trigram language model,In ICSLP-1996,2075-2078.
[121]Stanley F.Chen,Kristie Seymore,Ronald Rosenfeld.Topic Adaptation for Language Modeling Using Unnormalized Exponential Models.In ICASSP-98,Seattle,Washington,1998.
[122]R.Kneser,et al.Semantic Clustering for Adaptive Language Modeling 1997IEEE International Conference on Acoustics,Speech,and Signal Processing Germany,Apr.21-24,1997,pp.779-782,vol.2,
[123]H.Suzuki,J.Gao.A comparative study on language model adaptation techniques using new evaluation metrics,Proc.HLT/EMNLP,2005.
[124]Iyer,R.,M.Ostendorf,H.Gish.1997.Using Out-of-Domain Data to Improve In-Domain Language Models.IEEE Signal Processing Letters,4-8:221-223.
[125]Bacchiani,M.,B.Roark.Unsupervised Language Model Adaptation.Proceedings of ICASSP,2003,pp.224-227.
[126]Bacchiani,M.,B.Roark,M.Saraclar.Language Model Adaptation with MAP Estimation and the Perceptron Algorithm.Proceedings of HLTNAACL,2004,pp.21-24.
[127]Roark,B.,M.Saraclar,M.Collins.Corrective Language Modeling for Large Vocabulary ASR with the Perceptron Algorithm.Proceedings of ICASSP,2004,749-752.
[128]Mark Gales.Discriminative Models for Speech Recognition.Cambridge University Engineering Department.2007.
[129]Woodland,P.,Povey,D.Large scale discriminative training for speech recognition.In:Proc.ISCA ITRW ASR2000.2000,pp.7-16.
[130]于浩.判别训练在语言模型中的应用.上海交通大学硕士学位论文.2005.
[131]M.Collins.Discriminative reranking for natural language parsing,Proc.ICML,2000.
[132]M.Collins.Discriminative training methods for hidden markov models:theory and experiments with perceptron Algorithms,Proc.EMNLP,2002.
[133]J.Gao,H.Yu,W.Yuan et al.Minimum sample risk methods for language modeling,Proc.HLT/EMNLP 2005.
[134]袁伟.最小化样本风险算法在语言建模中的应用.上海交通大学硕士学位论文.2005.
[135]Zhengyu Zhou,Jianfeng Gao,Frank K.Soong,etc.A Comparative Study of Discriminative Methods for Reranking LVCSR N-Best Hypotheses in Domain Adaptation and Generalization.ICASSP 2006:141-144.
[136]R.Zhang,A.I.Rudnicky.Apply N-best list re-ranking to acoustic model combinations of Boosting Training,Proc.ICSLP,2004.
[137]C.Meyer.Utterance-level boosting of HMM speech recognizers,Proc.ICASSP,2002.
[138]Jen-Wei Kuo,Shih-Hung Liu,Hsin-Min Wang,etc.An Empirical Study of Word Error Minimization Approaches for Mandarin Large Vocabulary Continuous Speech Recognition.Computational Linguistics and Chinese Language Processing Vol.11,No.3,September 2006,pp.201-222.
[139]C.Allauzen,M.Mohri,B.Roark.Generalized algorithms for constructing language models.In Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics,pages 40-47,2003.
[140]M.Mohri,F.C.N.Pereira,M.Riley.The design principles of a weighted finite-state transducer library.Theoretical Computer Science,231:17-32,January 2000.
