摘要
字素音素转换是德语自然语言处理中的难点之一。提出一种基于决策树的字素音素转换的监督学习算法。在一个字素音素平行语料库的基础上,通过决策树进行字素音素转换的监督学习,生成字素音素转换规则。经交叉测试,平均转换正确率可达98. 03%。
The conversion of grapheme to phoneme is one of the difficulties in German natural language processing.In this paper,we proposed a supervised learning algorithm based on decision tree for grapheme-to-phoneme conversion.On the basis of a parallel corpus of grapheme and phoneme,the rules of morpheme-to-phoneme conversion were generated by supervised learning of grapheme-to-phoneme conversion through decision tree.Through cross test,the average conversion accuracy can reach 98.03%.
作者
王永生
李立贵
Wang Yongsheng;Li Ligui(School of Foreign Languages,Tongji University,Shanghai 200092,China)
出处
《计算机应用与软件》
北大核心
2019年第1期211-215,共5页
Computer Applications and Software
基金
同济大学中央高校基本科研业务费资助项目(1100219123)
关键词
德语字素音素转换
自然语言处理
决策树
German grapheme-to-phoneme conversion
Natural language processing
Decision tree
