Automatic Labeling of Voiced Consonants for Morphological Analysis of Modern Japanese Literature

摘要

Since the present-day Japanese use of voiced consonant mark had established in the Meiji Era, modern Japanese literary text written in the Meiji Era often lacks compulsory voiced consonant marks. This deteriorates the performance of morphological analyzers using ordinary dictionary. In this paper, we propose an approach for automatic labeling of voiced consonant marks for modern literary Japanese. We formulate the task into a binary classification problem. Our pointwise prediction method uses as its feature set only surface information about the surrounding character strings. As a consequence, training corpus is easy to obtain and maintain because we can exploit a partially annotated corpus for learning. We compared our proposed method as a preprocessing step for morphological analysis with a dictionary-based approach, and confirmed that pointwise prediction outperforms dictionary-based approach by a large margin.