Construction of Associative Vocabulary Learning System for Japanese Learners

Construction of Associative Vocabulary Learning System for Japanese Learners Takehiro Teraoka Faculty of Engineering, Takushoku University 815-1, Tatemachi, Hachioji-shi, Tokyo 193-0985, Japan tteraoka@cs.takushoku-u.ac.jp Tetsuo Yamashita Institute of Japanese Language Education, Takushoku University 3-4-14, Kohinata, Bunkyo-ku, Tokyo 112-8585, Japan tyamashi@ner.takushoku-u.ac.jp Abstract In recent years, natural language processing (NLP) technology has been applied to support systems and teaching materials for learning Japanese in the second language acquisition. Research on Japanese language learning using language resources and techniques related to NLP is divided into two main types: “dictio- nary/corpus” and “learning-support system”. While some corpora and learning-support sys- tems are useful, most relay on word co- occurrence information. To improve learning efficiency with a different approach, we aim to enable second language learners of Japanese to acquire vocabulary along with associative information of Japanese native speakers by using the Associative Concept Dictionary for Verbs (Verb-ACD). In this study, we have constructed a vocabulary learning system that generates question and answer sets and other correct candidate sets by extracting and com- bining associative information from the Verb- ACD. We investigated how accurately the pro- posed system could generate the question and answer sets for Japanese learners and found that it had the generation accuracy (0.70). We conclude that the various associative relation- ships among the generated words are impor- tant and require co-occurrence information to check their consistencies. 1 Introduction The number of Japanese language learners in the world has increased from about 580,000 in 1984 to approximately 3.67 million in 2015. While this number decreased a little from 2012 to 2015, the number of institutions for Japanese language edu- cation and the number of teachers increased slightly in a total of 137 areas including 130 countries and seven regions. These numbers cover only “school and other institutions teaching the Japanese lan- guage as language education”, and learners who study Japanese at other institutions or who self-study through various media (television, radio, books, the Internet, etc.) are not included. All told, the true number of Japanese language learners is estimated to be much higher (The Japan Foundation, 2017). Recently, natural language processing (NLP) technology has been applied to systems for learn- ing languages and to teaching materials for acquir- ing language knowledge. Research on Japanese language acquisition with language resources and techniques related to NLP is divided mainly into “dictionary/corpus” and “learning-support system” categories. The former includes the Japanese Learner’s Written Composition Corpus1 (Lee et al., 2013), which summarizes the composition data of Japanese learners, the Japanese Educational Vo- cabulary (Sunakawa et al., 2012), which contains 18,000 words for Japanese vocabulary education based on the Balanced Corpus of Contemporary Written Japanese (BCCWJ) (Maekawa et al., 2014) and a Japanese textbook corpus, and the Tsukuba Web Corpus (TWC)2, which includes 1,138 million words collected from the Web. The latter category includes the Learning Item Analysis System3, which outputs learning-item and 1http://sakubun.jpn.org 2http://nlt.tsukuba.lagoinst.info/ 3http://www.intersc.tsukuba.ac.jp/ kyoten/en/lias.html level-judgment results for input Japanese texts, and the text readability measurement system JReadabil- ity4, (Hasebe and Lee, 2015) which outputs the …

Content truncated to 500 words