Original-Transcribed Text Alignment for Manyosyu Written by Old Japanese Language

We are constructing an annotated diachronic corpora of the Japanese language. In part of this work, we construct a corpus of Man’yōsyū, which is an old Japanese poetry anthology. In this paper, we describe how to align the transcribed text and its original text semiautomatically to be able to cross-reference them in our Man’yōsyū corpus. Although we align the original characters to the transcribed words manually, we preliminarily align the transcribed and original characters by using an unsupervised automatic alignment technique of statistical machine translation to alleviate the work. We found that automatic alignment achieves an F1-measure of 0.83; thus, each poem has 1–2 alignment errors. However, finding these errors and modifying them are less workintensive and more efficient than fully manual annotation. The alignment probabilities can be utilized in this modification. Moreover, we found that we can locate the uncertain transcriptions in our corpus and compare them to other transcriptions, by using the alignment probabilities.