AI 기기를 활용한 한국어 발음 실태 조사 방법론 제안

임현열 한국문화융합학회 2019 문화와 융합 Vol.41 No.6

In this study, we propose a methodology for surveying Korean pronunciation using AI devices. Pronunciation surveys are necessary in the theoretical study of language, as such survey data are an important reference not in only phonology but also sociolinguistic research. In addition, these surveys are important in developing and improving language standard regulations after careful consideration of actual language use in speech. The necessity and importance of the pronunciation survey has been recognized in the past, and the survey has been conducted in various ways. However, past approaches were clearly limited in terms of research methodology. Therefore, this study proposes a new methodology to investigate the pronunciation of Korean language, using AI devices. This methodology is expected to have the following advantages: 1) suitable for collecting realistic pronunciations, 2) possibility of a large-scale pronunciation survey, 3) possibility of a real-time survey, 4) easy-to-understand characteristics of the pronunciation data provider, and 5) convenience in using the survey results in the research and policy-making process. At present, this survey methodology has not been implemented elsewhere and remains just a suggestion of a Korean linguist. It is hoped that this methodology of Korean pronunciation survey using AI devices will be converted into a real system through close collaboration between Korean linguists and engineers.

한국어 TTS에 적용되는 /ㄴㄹ/ 연쇄의 발음 처리에 대하여-TTS의 발음에 대한 국어 음운론적 관점의 진단-

임현열 국어국문학회 2019 국어국문학 Vol.0 No.187

이 연구에서는 한국어 TTS가 산출하고 있는 /ㄴㄹ/ 연쇄와 관련된 발음의 문제점을 살펴보고, /ㄴㄹ/ 연쇄의 적절한 발음 처리를 위해 TTS 시스템에서 고려해야 할 점들을 국어 음운론의 관점에서 제시하였다. 이를 위해 이 연구에서는 네이버 파파고와 구글 번역기의 TTS에서 다양한 /ㄴㄹ/ 연쇄를 어떻게 처리하고 있는지 살폈다. 그 결과, 한국어 TTS에서 /ㄴㄹ/의 유음화는 비교적 정확하게 처리하고 있는 데 반해, /ㄴㄹ/의 비음화와 관련하여 정확성이 낮다는 사실을 알 수 있었다. 또한, 이러한 현상의 원인을 한국어 TTS가 /ㄴㄹ/ 연쇄의 유음화와 비음화에 관여하는 형태소 정보를 적절히 반영하지 못하고 있기 때문이라고 진단하였다. 이에 이 연구에서는 한국어 TTS의 /ㄴㄹ/ 연쇄 발음 성능 향상을 위해 문자열을 발음열로 변환하는 과정에 형태소 정보를 반영하여 /ㄴㄹ/ 연쇄의 발음을 처리하는 방안을 제시하였다. 더불어, 위와 같은 방식으로 자연스러운 발음이 산출되기 어려운 일부 사례가 있기 때문에 해당되는 사례는 별도로 목록화하여 해당 단어의 발음 정보에 /ㄴㄹ/ 연쇄의 발음 처리 정보를 탑재해 두어야 한다는 점도 언급하였다. In this study, we investigated the pronunciation problems of /n-l/ string in Korean TTS system, and suggested the points to consider in TTS system for proper pronunciation processing of Korean TTS. To this end, this study examined how Naver Papago and Google Translator handle various /n-l/ strings. As a result, it was found that the accuracy of the lateralization of /n-l/ string was comparably high, but the accuracy of nasalization of /n-l/ string was very low. Also, it is diagnosed that the cause of this phenomenon is that the Korean TTS does not adequately reflect the morphological information related to the lateralization and nasalization of /n-l/ strings. In this study, we propose a method to process the pronunciation of /n-l/ string by reflecting the morpheme information in the process of converting the string to the pronunciation string in order to improve the pronunciation performance of Korean TTS. In addition, there are some cases where it is difficult to produce natural pronunciation in the above-mentioned method. Therefore, it is noted that the corresponding case should be listed separately and the pronunciation information of /n-l/ string should be loaded in the pronunciation information of the corresponding word.

ㅂ 불규칙 활용의 다차원

임현열 현대문법학회 2022 현대문법연구 Vol.113 No.-

ㅂ irregular inflexion is a phenomenon that occurs very frequently in Korean. Many researchers have been interested in this because the phenomenon is not simple. As a result, ㅂ irregular inflexion has long been the subject of interest with various people such as phonology, morphology, and syntax. Unlike this situation, language regul- ations and grammar textbooks treat ㅂ irregular inflexion too simply. It is difficult to fully understand the multidimensions of ㅂ irregular inflexion only with the content of the grammar textbook and the language regulations. There- fore, this study pointed out the problems of grammar textbooks and language regulations related to ㅂ irregular inflexion. In addition, various research results on the characteristics of ㅂ irregular inflexion were reviewed. Finally, all words listed in the Open Dictionary, <Urimalsaem> were reviewed. Through this, we tried to get closer to the practical knowledge of ㅂ irregular inflexion. This paper is expected to provide meaningful information for Korean language education and Korean language education as a foreign language in the future. .

한국어 TTS 경음화 발음의 국어학적 점검

임현열 한말연구학회 2019 한말연구 Vol.- No.54

The purpose of this study is to suggest how to improve pronunciation of Korean TTS from the point of view of Korean language. In particular, this study focused on the tensification pronunciation of TTS. For this purpose, this study examined the tensification pronunciation of Korean TTS. The pronunciation examination was done in two steps. The first step was a basic examination. We examined whether Korean TTS applied the six tensification rules of the standard pronunciation regulation. The six tensification rules were examined in 11 subtypes. The second step was a deepening examination. The problems arising from the basic examination were examined more closely. The deepening examination extended the number of words to be examined. Examination of the pronunciation of the TTS revealed problems with the four types of the tensification. Based on the results, this study proposed the following to improve the pronunciation of the tensification. (1) The pronunciation information must be faithfully attached to the words related to the tensification. (2) In order to process the tensification that occurs when the stem and the ending combine, the pronunciation information related to the tensification must be attached to the stem and the ending. (3) When the dependent nouns are combined after the adnominal form '-(으) ㄹ' in predicate, the rules of tensification should be applied carefully. (4) Regardless of the spacing, when a chain of obstruent sounds occurs in consecutive nouns, the obstruent sound in the back should be pronounced as tensificated sound. In order for these studies to lead to substantial improvements in TTS, collaboration between engineering and linguistics is required.

한국어 교재 어미 출현 분포의 정보학

임현열 현대문법학회 2023 현대문법연구 Vol.119 No.-

. The purpose of this study is to find and discuss linguistics implications that can be found in the distribution of endings in Korean textbooks through quantitative analysis of listening and reading passages in Korean textbooks. Through this work, we will have more meaningful information about the distributional characteristics of the appearance of endings in Korean textbooks, and it is expected that this information can be used as an objective basis for seeking improvements in Korean textbooks. This study used Korean textbooks published by Seoul National University as analysis data. Python was used as an analysis tool. The kkma morpheme analyzer included in the konlpy package was used. For analysis, functions such as ‘kkma.sentences()’, ‘kkma.morphos()’, ‘kkma.pos()’, and ‘len()’ were applied. I went through the process of configuring it as a data frame using the pandas library. Through this process, it was possible to know the characteristics of the type frequency and appearance frequency of final endings, pre-final endings, conjunctive endings, and function-converting endings in listening and reading passages, respectively. In addition, the necessary items for the kkma morpheme analyzer to perform more appropriate endings analysis were also mentioned.

[K-문화 소식] 경성대학교 한국한자연구소, 대중과 함께하는 지역인문학센터

임현열 K-문화융합협회 2023 K-문화융합저널 Vol.3 No.1

한국어 음성합성 기술의 초분절음 진단 체계: 언어치료용 AI에 적용될 기술을 예시로 하여

임현열 강원대학교 인문과학연구소 2023 인문과학연구 Vol.- No.78

This study was initiated to find ways to prevent speech synthesis technology from causing the uncanny valley effect to users in the field of speech therapy. Through past research, microdiagnosis of segmental sounds and solutions for them were explored. Here, the items to be diagnosed and the diagnostic methods for microdiagnosis of suprasegmental sounds were systematically presented. The diagnostic items for suprasegmental sounds were categorized into three dimensions: length, pitch, and intensity of the sound. For each dimension, it has been shown that pause duration and speech rate should be considered as diagnostic criteria for 'length,' pitch range, pitch level, and pitch variability for 'pitch,' and sound pressure level, peak amplitude, mean square amplitude, and other factors for 'intensity.' Moreover, these measurements, in common, are not solely concerned with achieving accurate speech; they are also relevant to the nuanced aspects of emotional expression, depending on the degree and level. Therefore, we meticulously examine the characteristics inherent in natural human speech for these elements and establish the current voice. It was mentioned that if you calmly reflect the elements that cannot be implemented in synthetic technology, you can enter a more complete technology section where you do not feel the uncanny valley. Nevertheless, it is essential to emphasize that the speech therapy AI discussed here serves as an exemplar of AI requiring delicate communication, and the scope of application proposed in this study can be extended to encompass psychological counseling AI, educational AI, and beyond.

南北 言語 統合 政策에 대한 專門家 調査 結果 分析

임현열,최유숙 한국어문교육연구회 2015 어문연구(語文硏究) Vol.43 No.3

This study aims to analyze the opinions of experts of various fields on the policies proposed for the linguistic unification of South and North Korea. This study collected policies designed for linguistic unification that were proposed since 2000. After reviewing the selected policies, a list containing 18 linguistic unification policies was finalized. Then a survey was conducted on experts of various fields. The survey provided information on the listed policies and asked the participants to grade the policy according to importance, urgency, and utility on a five point scale. The results were analyzed based on the area of expertise of the participant and the groups to which they belong. The results are expected to be meaningful since it will serve as a cornerstone for future policy implementation. 이 硏究의 目的은 南北 言語 統合을 爲해 提案된 여러 政策들에 對한 各界 專門家의 見解를 分析하는 데 있다. 이 硏究에서는 2000年 以後 提案된 南北 言語 統合에 關한 多樣한 政策을 蒐集하였다. 蒐集된 政策을 檢討하여 最終的으로 18個 政策이 包含된 南北 言語 統合 政策 目錄을 導出하였다. 政策 目錄에 包含된 各 政策에 對해 各界 專門家들에게 設問 調査를 實施하였다. 設問 調査는 政策 目錄과 各各에 對한 說明을 提示한 後, 各 政策에 對한 重要度, 時急度, 活用度를 5段階 評價 尺度에서 評價하게 하는 方式으로 進行되었다. 設問 調査 結果는 專門家의 專攻 領域과 所屬 集團의 特性에 따라 區分하여 分析하였다. 이러한 調査 및 分析 結果는 向後 南北 言語 統合 政策 執行에 있어 重要한 判斷 根據가 될 수 있다는 點에서 意味가 있을 것으로 豫想된다.

조사 분포의 계량적 측정을 통한 정보 탐색 연구― 구어성 텍스트와 문어성 텍스트의 특성 비교 ―

임현열 중앙어문학회 2023 語文論集 Vol.95 No.-

This study aims to quantitatively analyze the distribution of particles(조사 [josa] in Korean linguistics) in Korean texts. Moreover, the study examines how particle usage reveals characteristics of spoken and written texts. The study investigates spoken and written language features in texts by comparing unique texts. These may include comparing reading and listening to passages in Korean textbooks with those in a balanced corpus. For this purpose, the “bareun” deep learning-based morphological analyzer, coupled with Python’s “konlpy” library, was employed as the analytical framework. Furthermore, the study used Excel’s pivot tables, various filter functions, and correlation coefficient functions to analyze the data from multiple perspectives. The results provide insight into the use of particles in the ongoing discourse on the characteristics of spoken and written Korean texts. Additionally, this study identified certain characteristics in Korean textbook texts, particularly in listening passages, which may not fully represent aspects of the Korean language.

『다문화 감수성을 기반으로 한 한국어 특화반 수업 연구』에 대한 토론문

임현열 국제어문학회 2016 국제어문학회 학술대회 자료집 Vol.2016 No.1