순차통역 보조 툴로서 자동음성인식 사용성의 사례연구

이주리애 인문사회 21 2022 인문사회 21 Vol.13 No.4

A Case Study on the Usability of Automatic Speech Recognitionas an Auxiliary Tool for Consecutive InterpretingJuriae Lee Abstract: This research aims to explore the usability of ASR as CAI. Three-trial experiments were conducted in order to allow interpreters to familiarize themselves with using ASR. The participants expressed their impressions on the usability of ASR through a questionnaire. They had grown a sense of trust in ASR after trying it. The level of satisfaction and eye gaze processing varied according to text length and content. Interpreters were observed to contemplate on how to manage their gaze and utilize the transcribed texts as trials were repeated, and preferred to be provided with texts transcribed by ASR for psychological stability. The findings could be significant in assessing the usability of ASR in practical interpretation at its current level. Follow-up studies are expected to extend to assessing the usability of real-time ASR assistance at off-line consecutive interpreting environments. Key Words: Consecutive Interpreting, Automatic Speech Recognition, Usability, Interpreter, Computer-assisted Interpreting (CAI) 순차통역 보조 툴로서 자동음성인식 사용성의 사례연구이 주 리 애** 연구 목적: 이 연구는 최근 정확도가 크게 향상된 자동음성인식의 순차통역 보조 툴 활용을 탐색하는 데 목적이 있다. 연구 방법: 전문통역사를 대상으로 실험한 후 설문조사를 통해 자동음성인식의 사용성과 활용 가능성 및 보완점을 질문했다. 본 연구는 자동음성인식이 익숙해지는 과정을 고려해 3회에 걸친 실험으로 설계해 진행했다. 연구 내용: 실험참여자는 대부분 실험을 통해 음성인식에 대한 신뢰를 갖게 되었다고 했다. 음성인식 사용의 만족도 및 시선처리는 텍스트 길이와 내용에 따라 차이를 보였다. 회차를 거듭하면서 시선처리 및 활용 방안을 고민하는 모습이 보였고, 심리적 안정을 위해 음성인식 제공을 대부분이 희망했다. 결론 및 제언: 본 연구의 결과는 실무 통역에서 현 수준의 음성인식 활용에 유의미할 것으로 사료되며, 향후 오프라인 순차통역 환경에서 실시간으로 제공되는 자동음성인식의 사용성 연구로 확장해 나가고자 한다. 핵심어: 순차통역, 자동음성인식, 사용성, 통역사, 통역 보조 도구 □ 접수일: 2022년 7월 11일, 수정일: 2022년 7월 26일, 게재확정일: 2022년 8월 20일* 이 논문은 2020년 대한민국 교육부와 한국연구재단의 인문사회분야 신진연구자 지원사업의 지원을 받아 수행된 연구임(NRF-2020S1A5A8047479). ** 이화여자대학교 통역번역대학원 부교수(Professor, Ewha Womans Univ., Email: wnfldo@hanmail.net)

자동음성인식의 통역 활용에 관한 예비연구

이주리애 ( Juriae Lee ) 사단법인 아시아문화학술원 2021 인문사회 21 Vol.12 No.5

본 논문은 자동음성인식의 통역 활용을 탐색하는 데 목적이 있다. 이를 위해 음성인식의 형태를 분류하고 통역번역대학원생을 대상으로 음성인식 활용 통역 실험 후 설문조사를 통해 음성인식의 통역 활용에 대한 수용을 논하였다. 분류한 다양한 형태의 스피치 투 텍스트를 활용한 통역 실험 결과, 순차통역과 동시통역 모두 음성인식이 어느 정도 도움이 되었다는 의견이 많았다. 향후 전망은 통역 보조 도구로서 음성인식 및 기계번역이 통역 현장에서 활성화될 것이며, 그렇다고 해도 순차통역할 때 노트테이킹은 필요하다는 의견이 많았다. 또한, 통역 교육에서 음성인식 활용이 필요하고 음성인식의 성능이 생각보다 좋았다고도 했다. 본 연구는 음성인식을 실제로 통역에 적용해 보고 그 사용성 탐구를 시도한 데 의의가 있으며, 향후 구체적인 음성인식의 활용 및 교육 방식을 제안할 필요성이 있음을 시사한다. The purpose of this study is to investigate the possible application of Automatic Speech Recognition in interpretation. To this end, ASR was categorized according to type and then tested by students attending graduate schools of interpretation and translation, who thereafter answered a questionnaire which results are inquired and discussed. The majority stated that ASR was helpful to a certain degree in both consecutive interpretation and simultaneous interpretation following a series of tests using categorized Speech-to-Text samples. Many believed that ASR and machine translation, such as Computer-assisted Interpreting Tool, would be actively utilized at interpreting sites in the future, however, most maintained that note taking would still be needed. This research is significant in that it actually utilizes ASR on actual interpretations in order to assess its applicability. The study also raises the need to propose specific ASR utilization and education methods.

디지털 환경 변화에 따른 통역 교육― 자동음성인식 활용을 중심으로 ―

이주리애 한국일본학회 2022 日本學報 Vol.- No.133

With changes in the digital environment, the technology used for Computer Assisted Interpretation (CAI) and Computer Assisted Interpretation Training (CAIT) has diversified and studies related to the use of technology for interpreting is increasing since 2020. This paper focuses on the use of Automatic Speech Recognition (ASR) for interpreting and interpreter training, which is an emerging field of study, by summarizing the current state and trend in terms of tools that can be used for interpreter training. The currently available all-purpose multilingual ASR applications (apps) were classified by function and they were then categorized according to whether they can be used for transcribing original texts or translated texts. Next, these were divided into apps with real-time ASR functions that can be used for interpretation training, and apps with recording and transcribing functions that can be used for post-interpretation correction and reflection. The real-time ASR apps were further categorized according to timeline display availability and the number of preset languages. The present work also introduces an automatic interpretation performance evaluation system prototype that is currently under development. It is a platform that extracts from ASR results every silence, filler and backtracking which occur during interpretation to provide processed statistics. It is expected that various ways for introducing ASR into interpretation education will be explored in the future, from beginner to advanced level, following sufficient practical adaption through training. 디지털 환경의 변화에 따라 컴퓨터 보조 통역(CAI)과 컴퓨터 보조 통역 교육(CAIT)에 활용되는 기술이 다양해진 가운데 한국과 일본도 2020년을 기점으로 기술의 활용과 관련된 통역 연구가 증가하고 있다. CAI 및 CAIT는 유럽을 중심으로 다양한 툴이 개발되고 있으며, 전문용어 관리 등 통역의 사전준비와 원격동시통역 시스템에서 음성인식으로 통역을 보조하기도 한다. 본 논문에서는 최근 부상하고 있는 자동음성인식(ASR)을 활용한 통역과 통역 교육의 현황을 조사하여 통역 교육에 활용할 수 있는 툴을 정리했다. 먼저 현재 사용 가능한 다국어 음성인식 범용 애플리케이션을 기능별로 분류했다. 그리고 원문과 통역문에 각각 활용할 수 있는 앱으로 나누어 제시했다. 다음으로 통역 실무에서 활용하기 위한 실시간 음성인식 앱, 통역 훈련에 사용 가능한 녹음 후 전사하는 기능의 앱으로도 나누었다. 실시간 음성인식 앱은 타임라인 표시의 유무, 사전에 설정해 놓을 수 있는 언어의 수에 따라 나뉜다. 녹음 후에 전사되는 앱은 통역 과제물을 첨삭하거나 자가 학습에 활용할 수 있다는 점에서 유용하다. 마지막으로 현재 개발 중인 통역 퍼포먼스의 자동평가 시스템의 프로토타입을 소개했다. ASR 결과로부터 통역에서 발생할 수 있는 침묵, 필러, 백트래킹을 추출해 통계 처리해 주는 시스템이다. 향후 음성인식을 통역 교육에 도입하는 방안이 다각도로 모색되어 훈련 과정에서 충분히 적응함으로써 초급에서 고급 및 실무에 이르기까지 유용하게 활용되기를 바란다.

다국어 자동 통번역을 위한 공통 변환 기반 하이브리드 자동 번역 방법

최승권,김영길 한국외국어대학교 통번역연구소 2016 통번역학연구 Vol.20 No.3

Along with the advancement of speech recognition technology and machine translation technology in addition to the fast distribution of mobile devices, automatic speech translation technology no longer remains as a subject of research as it has become popularized throughout many users. The goal of this paper is to describe the method of hybrid machine translation based on common transfer for multilingual speech and machine translation. As of 2012, there are different bilingual machine translation systems such as Korean-to-English, English-to-Korean, Korean-to-Japanese, Japanese-to-Korean, Korean-to-Chinese, and Chinese-to-Korean machine translation systems in Korea. An increase in the number of tourists is expected due to the fact that the Winter Olympic Games will be held in Korea in 2018. Budget problem prevents us from employing a sufficient number of human interpreters who can fluently interpret such languages as Spanish, French, German, and Russian. For this reason, consumers such as the Korea Tourism Organization and speech translation service providers demand a multilingual speech and machine translation system. We are developing a multilingual speech and machine translation system aiming for multilingual expansibility and high-quality translation accuracy.

정보기술을 활용한 중국 통역 교육 연구 현황

손지봉,김천향,공수 이화여자대학교 통역번역연구소 2024 T&I review Vol.14 No.1

Jibong Son, Tianxiang Jin and Su Kong (2024). Current status of Chinese interpretation education using information technology. This paper is a research study on the current status of interpretation education in China using information technology. The study was divided into five categories: developmental history, recognition of the interpretation corpus, a new interpretation education model, evaluation of the standard of automatic interpretation ability, and the use of voice recognition. First, China is shifting to the stage of convergence between information technology and interpretation education. Second, the country is seeking to build an interpretation education corpus. Third, it is seeking a hybrid learning and flip learning interpretation education model equipped with an online platform. Fourth, there is a consensus on the usefulness of automatic evaluation. Fifth, hand gestures, as well as sub-languages that consist of sound other than language, have reached the level of speech recognition that can be displayed with transcription and annotation. In particular, the implication can be derived that systematic support from the state was at the root of China's development in this field. (Ewha Womans University, Korea)

한-영 실시간 자동통역 시연 사례 고찰

최은실,김아영 한국통번역교육학회 2020 통번역교육연구 Vol.18 No.2

This research aims to find out the principal of Machine Interpretation (MI) and the difference of interpreting performed by human and machine. Also, it identifies how the real-time Machine Interpretation system is operated and analyses the interpretation quality assessment. First, the MI consists of process of speech recognition, machine translation and speech synthesis, while human interpretation consists of the process of listening and reexpression. In the case of the ETRI’s Real-time Automatic Speech Translation Model, it replaces the process of speech synthesis with that of subtitle translation. And after establishing the learning data on output from machine translation, it demonstrates its function in a real class with a professor and students. Based on source/target texts and a quality assessment sheet, the study finds three characteristics of Machine Interpretation: 1) it provides wrong messages caused by difficult equipment control, 2) low simultaneity because of the difference in transfer units, and 3) low coverage because of the difference between a verbal interpreting and written subtitle translation.

통역사의 어휘전환 자동화

정혜연(Chung, Hye-yeon) 한국외국어교육학회 2013 Foreign languages education Vol.20 No.3

This paper examines the development of lexical transfer during interpreting training, more precisely, whether one's competence to translate a word into another language improves to its automaticity while acquiring interpreting competence (Chung, 2012). According to Bargh (1992) the state of automaticity can be broadly defined by four features; efficiency, awareness, intention and control. Due to the cognitive complexity of lexical transfer; however, it cannot be performed in a state of unawarenesss or unintentionally. Nor can it be in an uncontrollable form, so that its automaticity is only to be measured by its efficiency. Sixteen subjects, divided into four groups according to their training years, took a test for lexical transfer efficiency. The result showed, in all three categories of efficiency (rapidity, its coefficient of variability (CV) and accuracy) the years of interpreting training and the efficiency correlated. Although in terms of rapidity, re-tests with a broader range of subjects are necessary for the results to be statistically insignificant. From these findings we can conclude that our hypothesis is, at least in part, confirmed. In der vorliegenden Arbeit geht es um die Entwicklung der lexikalen Kompetenz bei Dolmetschstudenten sowie professionellen Dolmetschern, genauer gesagt, wie sich bei ihnen die Kompetenz der lexikalen Transfer(von der Ausgangssprache in die Zielsprache) kontinuierlich entwickelt. Zu überpüfen ist dabei die Hypothese. Automatisierung der lexikalen Transfer schreitet fort, je länger einer Dolmetschkompetenz erwirbt(Chung, 2012). Automatisierung einer Handlung läßt sich laut Bargh(1992) dadurch erkennen, dass die betreffende Handlung effizient, nicht-intentional, unbewußt und unkontrollierbar durchgeführt wird. Da lexikale Transfer aufgrund ihrer kognitiven Komplexität weder nicht-intentional, noch unbewußt auszuführen ist, wird die Automatisierung dieser Aufgabe an deren Effizienz gemessen. 16 Versuchspersonen, die nach ihrer Dolmetschausbildungszeit in vier Gruppen geteilt sind, werden einem solchen Effizienztest unterzogen. In allen drei Kategorien der Effizienz(Schnelligkeit, Variabilitätskoeffizient (CV) und Genauigkeit) korrellierte die Ausbildungszeit mit den Effizienzsmesswerten - wobei der Unterschied in Schnelligkeit statistisch nicht signifikant war -, so dass wir zu dem Schluß kamen, dass die Hypothese zumindest teilweise zu bestätigen ist.

Limitations of Machine Translation : Focused on the Role of Cognitive Complements in Human Communication

Jin, Sil-Hee 한국외국어대학교 통번역연구소 2017 통번역학연구 Vol.21 No.4

Significant progress has been made in machine translation since the introduction of the Neural-based Machine Translation (NMT) algorithms. For Korean-English translation, there is growing evidence that the machine translation engines are rendering more and more acceptable translations with time through deep learning. The progress inevitably raises the question of whether human expertise will eventually be replaced by machines. This question depends on whether and how much of the inter-lingual mediation process depends on inherently human capabilities. To understand the potential boundaries of machines translation, this study draws on the role of cognitive complements in face-to-face communication, focusing on spoken as opposed to the commonly researched written texts. The first part of the study discusses the role of cognitive complements in human communication. Then to illustrate the important role played by cognitive complements in human communication, the second part of the study explains the errors made by NMT from the perspective of cognitive complements, based on a sample text analysis using Google Translate. The analysis indicate that applying the current engine to face-to-face encounters has serious pragmatic risks. From the perspective of machine translation, therefore, the ability to apply cognitive complements like human beings, with the ability to draw on extra-linguistic context may be the key to making NMT ready for real-world applications as far as the processing of spoken texts is concerned. While effort is underway to relate sensory experience to text processing, enormous computing power requirement may be another obstacle before human interpreters can be replaced completely. Pedagogic implications of this study for post-graduate interpreting programs are discussed in the final part of the paper.

한국어 음성데이터를 이용한 일본어 음향모델 성능 개선

이민규,김상훈,Lee, Minkyu,Kim, Sanghun 한국음향학회 2013 韓國音響學會誌 Vol.32 No.5

본 논문은 일본어 음성인식기 신규 개발을 위해 초기에 부족한 일본어 음성데이터를 보완하는 방법이다. 일본어 발음과 한국어 발음이 유사한 특성을 근거로 한국어 음성 데이터를 이용한 일본어 음향모델 성능개선 방법에 대하여 기술하였다. 이종언어 간 음성 데이터를 섞어서 훈련하는 방법인 Cross-Language Transfer, Cross-Language Adaptation, Data Pooling Approach등 방법을 설명하고, 각 방법들의 시뮬레이션을 통해 현재 보유하고 있는 일본어 음성데이터 양에 적절한 방법을 선정하였다. 기존의 방법들은 훈련용 음성데이터가 크게 부족한 환경에서의 효과는 검증되었으나, 목적 언어의 데이터가 어느 정도 확보된 상태에서는 성능 개선 효과가 미비하였다. 그러나 Data Pooling Approach의 훈련과정 중 Tyied-List를 목적 언어로만으로 구성 하였을 때, ERR(Error Reduction Rate)이 12.8 %로 성능이 향상됨을 확인하였다. In this paper, we propose an enhancement of Japanese acoustic model which is trained with Korean speech database by using several combination strategies. We describe the strategies for training more than two language combination, which are Cross-Language Transfer, Cross-Language Adaptation, and Data Pooling Approach. We simulated those strategies and found a proper method for our current Japanese database. Existing combination strategies are generally verified for under-resourced Language environments, but when the speech database is not fully under-resourced, those strategies have been confirmed inappropriate. We made tyied-list with only object-language on Data Pooling Approach training process. As the result, we found the ERR of the acoustic model to be 12.8 %.