基于贪婪算法的俄语字音快速转换方法  

Fast Conversion Method of Russian Characters Based on Greedy Algorithm

作　　者：王文娇 WANG Wen-jiao(Weinan Normal University,Weinan 714099 China)

机构地区：[1]渭南师范学院,陕西渭南714099

出　　处：《自动化技术与应用》2022年第10期77-80,共4页Techniques of Automation and Applications

摘　　要：为高效完成俄语字音转换,准确处理大规模俄语文本问题,提出基于贪婪算法的俄语字音快速转换方法。将字音转换看作统计分类问题,通过确定俄语特征与约束条件、最大熵准则与指数,构建最大熵多音词消歧模型;将俄语词典表述为马尔科夫模型,采用规则与统计相结合的语言模型评价所有转换路径,建立语言分析树,根据设立的多种假设及评价函数,得出最优转换路径;利用贪婪算法添加规整因子,将语音信号设置为适当长度,完成字音快速转换。实验结果表明,所提方法输出的俄语语音信号较为平稳,字音转换中降低了信号丢失以及转换的错误率。In order to efficiently complete the Russian word to speech conversion and accurately deal with large-scale Russian text problems, a fast Russian word to speech conversion method based on greedy algorithm is proposed. This paper regards the word sound conversion as a statistical classification problem, and constructs the maximum entropy multi tone word disambiguation model by determining the Russian features and constraints, maximum entropy criterion and index;the Russian dictionary is expressed as a Markov model, and the language model combining rules and statistics is used to evaluate all the conversion paths.The language analysis tree is established, and the optimal conversion path is obtained according to various assumptions and evaluation functions;the greedy algorithm is used to add the regularization factor, and the speech signal is set to the appropriate length to complete the fast conversion. The experimental results show that the Russian speech signal output by the proposed method is relatively stable, and the signal loss and error rate of conversion are reduced.

关 键 词：贪婪算法 俄语字音转换 最大熵模型 规整因子 评价函数 马尔科夫模型 

分 类 号：TP318.2[自动化与计算机技术—计算机软件与理论]