Compilation of “Topic-Vocabulary for Japanese Language Education” from a Corpus of Natural Conversation

Abstract

This article describes how to create and use the “Topic-Vocabulary Table for Japanese Language Education” which was elaborated based on a conversation corpus. To create this table, several workers manually checked and divided the “Nagoya University Conversation Corpus” into subcorpora for each topic. Then, log-likelihood ratios were calculated for each sub-corpus. Finally, an Excel table was created with 97 topics arranged horizontally and 3,324 words arranged vertically. This table can be used for Japanese language education in two directions: from topic to word, and from word to topic. In the former, users can learn function words and language behaviors in addition to frequently used words in the topic. In the latter, users can notice that synonyms are frequently used in different topics for each, as well as a bias in the topics in which functional words are used.