俄汉平行语音语料库词性标注一致性检查方法

打开文本图片集

关键词：俄汉平行语音语料库；语音识别；词性标注；隐马尔可夫模型；双向循环神经网络；一致性检查中图分类号：TN911.6-34；TP183 文献标识码：A 文章编号：1004-373X（2025）20-0142-05

Method of part-of-speech tagging consistency check in Russian-Chinese parallel speech corpus

CHONG Huifang （National UniversityofDefense Technology，Nanjing21Oo39，China）

Abstract：Russian-Chineseparalel speechcorpushasextensiveapplicationvalueinlanguageresearch，cross-cultural communication，languageengineeringandotherfields.Onthisbasis，amethodof parts-of-speechtagingconsistencycheck in Rusian-Chineseparalelspeechcorpusisproposed toensuretheaccuracyanduniformityofparts-of-speech taging inthe corpus，improvethequalityofparts-of-speech tagging inRusian-Chineseparalelspeechcorpus，improvetheproceing eficiencyof Rusian-Chinesebilingual task，andprovideareliabledata basisforsubsequent languageresearch，machine translationandotherapplications.ThebidirectionalrecurrntneuralnetworkisusedtoidentifytheRusian-Chinesepaalel speechinthecorpus，andconvertthespeechdataintotextdatatoformtheRussian-Chinese paraleltext.Thesetof part-ofspechtaggingofRussan-ChineseparaleltextsisobtainedbymeansofhiddenMarkovmodel，andtheoptialpartof-speech taggingsequenceissolvedbymeansofViterbialgoritm，sastorealizethepartof-speechtagingofRuia-Chiesepaalel texts.Accordingtotheclusteringidea，thebarycentricclusteringalgorithmisusedtojudgewhetherthepart-of-speechtaggingof Rusian-Chineseparaleltexts isonsistent，soastorealiethecosistencycheckofpart-of-spechtaggingofRusian-Chiese paralelspeechcorpus.Theexperimentalresultsshowthattheproposedmethodcanefectivelyrealizetheconsistencycheckof part-of-speech tagging in Russian-Chinese parallel speech corpus，and has good accuracy and reliability.

Keywords：Russian-Chinese paralel speech corpus;speech recognition;part-of-speech tagging; Hidden Markovmodel; bidirectional recurrent neural network;consistency check

0 引言

俄汉平行语料库作为语言资源的重要组成部分，为俄汉语言对比研究、机器翻译、自然语言理解等提供了丰富的数据支持[1-2]。（剩余5948字）