A Study of Japanese Vocabulary Recognition Teaching Strategies Based on Deep Belief Networks

Abstract

Abstract Based on the deep belief network, this paper proposes a teaching strategy for Japanese vocabulary recognition with the main direction of realizing scientific classification of Japanese vocabulary and building a corpus. In order to improve the classification effect of Japanese vocabulary, a Japanese vocabulary classification model based on a deep belief network is constructed, and the TF-IDF and LSI features of the text are used as inputs to obtain the deep-level features and improve the final classification effect. A Japanese vocabulary corpus is constructed through the steps of corpus collection, database design, and corpus production, and a Markov number labeling model based on a deep belief network is constructed to transform the problem of sentence tense processing into the problem of labeling tense trees and propose a DBN-based method for translating the tense of the corpus. The teaching practice of Japanese vocabulary recognition was conducted with 100 first-year students at the College of Japanese Language at Dalian University of Foreign Languages as research subjects. The experimental class scored slightly higher than the control class in all six Japanese vocabulary tests, and the mean values of vocabulary utilization ability and vocabulary comprehension ability were higher than those of the control class by 2.94 and 1.06, respectively, which showed a significant difference (P<0.05). The experimental class also scored higher than the control class in all dimensions of Japanese vocabulary learning ability, showing significant differences (P<0.05).