Affiliations: [a] Department of Convergence Software, Hallym University, 1, Hallymdaehak-gil, Chuncheon-si, Gangwon-do, Republic of Korea
| [b] Bio-IT Center, Hallym University, 1, Hallymdaehak-gil, Chuncheon-si, Gangwon-do, Republic of Korea
| [c] Division of Speech Pathology and Audiology, Hallym University, 1, Hallymdaehak-gil, Chuncheon-si, Gangwon-do, Republic of Korea
Correspondence:
[*]
Corresponding author. Yu-Seop Kim, E-mail: yskim01@hallym.ac.kr.
Abstract: Speech pathology is a scientific study of speech disorders. In this field, the study also analyzes and evaluates language abilities for the purpose of improving speech and hearing. Speech therapy first performs evaluation of speech ability, which is expensive. In order to solve this problem, software methodologies have been applied to language analysis, but most of them have been applied to only part of the whole process. In this study, the degree of language development is judged by determining the age group of the speaker (Pre-school children, Elementary school, Middle and high school, Adults, and Senior citizen) using deep learning and simple statistics. We use transcription data from the counseling contents and multi-kernel CNN model. At this time, in order to understand the characteristics of Korean language belonging agglutinative languages, experiments are carried out in words, morphemes, characters, Jamo, and Jamo with POS tag-level. And we analyze the distribution of the results for each sentence of the speakers to predict their age groups and to check the degree of language development. The proposed model shows an average accuracy of about 74.6 %.
Keywords: Language analysis, age group analysis, convolutional neural networks, deep
learning, statistical analysis
