德语字素音素转换学习语料库的自动生成算法

Automatic generation algorithm of German grapheme-to-phoneme conversion learning corpus

提出了一种基于迭代的有限状态转录机的字素音素转换学习语料库的自动生成算法,针对一个德语单词及其音标进行基于形态规则和语音规则的迭代,最终将字素和音素完全一一对应起来。经过抽样测试,算法的正确率可以达到98.6%。

An algorithm based on iterated finite state transducer was proposed to automatically create the grapheme-to-phoneme conversion learning corpus. In the algorithm, graphemes in a word and phonemes in its IPA will correspond after iterations based on morphological rifles and phonological rifles. The accuracy of sampling test reached 98.6%.