• The Transactions of the Korea Information Processing Society
• /
• v.13 no.3
• /
• pp.140-147
• /
• 2024

In the Republic of Korea, people with hearing impairments are the second-largest demographic within the registered disability community, following those with physical disabilities. Despite this demographic significance, research on sign language translation technology is limited due to several reasons including the limited market size and the lack of adequately annotated datasets. Despite the difficulties, a few researchers continue to improve the performacne of sign language translation technologies by employing the recent advance of deep learning, for example, the transformer architecture, as the transformer-based models have demonstrated noteworthy performance in tasks such as action recognition and video classification. This study focuses on enhancing the recognition performance of sign language translation by combining transformers with 3D-CNN. Through experimental evaluations using the PHOENIX-Wether-2014T dataset [1], we show that the proposed model exhibits comparable performance to existing models in terms of Floating Point Operations Per Second (FLOPs).

• Journal of Digital Convergence
• /
• v.18 no.6
• /
• pp.271-277
• /
• 2020

Machine translation refers to software that translates a source language into a target language, and has been actively researching Neural Machine Translation through rule-based and statistical-based machine translation. One of the important factors in the Neural Machine Translation is to extract high quality parallel corpus, which has not been easy to find high quality parallel corpus of Korean language pairs. Recently, the AI HUB of the National Information Society Agency(NIA) unveiled a high-quality 1.6 million sentences Korean-English parallel corpus. This paper attempts to verify the quality of each data through performance comparison with the data published by AI Hub and OpenSubtitles, the most popular Korean-English parallel corpus. As test data, objectivity was secured by using test set published by IWSLT, official test set for Korean-English machine translation. Experimental results show better performance than the existing papers tested with the same test set, and this shows the importance of high quality data.

• Journal of Korea Multimedia Society
• /
• v.8 no.6
• /
• pp.790-797
• /
• 2005

It is well-known that there are many similarities between Japanese and Korean language. For example, the order of words and the nature of the grammatical conjugation of both languages are almost the same. Another similarity is the frequent omission of the subject from a sentence. Moreover, both languages have honorific expressions and the identical concept for expressing nouns in terms of Chinese characters. Using these similarities, we have developed a word-to-word translation system which does away with any deep level analysis of syntactic and semantic structures of the two languages. If we use these similarities, the direct translation method is superior to the internal language translation method or transfer-based translation method. Although the MT system based on the direct translation method is more easily developed than the ones based on other methods, it may have a lot of difficulties when it tries to select the appropriate target word from ambiguous source verbs. In this paper, we propose a new algorithm to extract the meaning of substantives and to make use of the order of the extracted meaning. We could select $86.5\%$ appropriate verbs in the sample sentences from IPAL-verb-dictionary. $13.5\%$ indicates the cases in which we could not distinguish the meaning of substantives. We are convinced, however, that the succeeding rate can be increased by getting rid of the meaning of verbs thatare not used so often.

• Annual Conference on Human and Language Technology
• /
• 2012.10a
• /
• pp.215-218
• /
• 2012

본 논문에서는 한국어 대화 처리를 위한 통계기반 음성언어이해 시스템에 대해 기술한다. 음성언어이해시스템은 대화처리에서 음성 인식된 문장으로부터 사용자의 의도를 인식하여 의미표현으로 표현하는 기능을 담당한다. 한국어의 특성을 반영한 실용적인 음성언어이해 시스템을 위해서 강건성과 적용성, 확장성 등이 요구된다. 이를 위해 본 시스템은 음성언어의 특성상 구조분석을 하지 않고, 마이닝 기법을 이용하여 사용자 의도 표현을 생성하는 방식을 취하고 있다. 또한 한국어에서 나타나는 특징들에 대한 처리를 위해 자질 추가 및 점규화 처리 등을 수행하였다. 정보서비스용 대화처리 시스템을 대상으로 개발되고 있고, 차량 정보서비스용 학습 코퍼스를 대상으로 실험을 하여 문장단위 정확률로 약 89%의 성능을 보이고 있다.

• IEIE Transactions on Smart Processing and Computing
• /
• v.5 no.2
• /
• pp.77-84
• /
• 2016

In the field of natural language processing, Universal Networking Language (UNL) has been used by various researchers as an inter-lingual approach to automatic machine translation. The UNL system consists of two main components, namely, EnConverter for converting text from a source language to UNL, and DeConverter for converting from UNL to a target language. Currently, many projects are researching how to apply UNL to different languages. In this paper, we introduce the tools that are UNL's applications and discuss how to reuse them to encode a Vietnamese sentence into UNL expressions and decode UNL expressions into a Vietnamese sentence. The testing was done with about 1,000 Vietnamese sentences (a dictionary that includes 4573 entries and 3161 rules). In addition, we compare the proportion of sentences translated based on a direct method (Google Translator) and another one based on UNL.

• Annual Conference on Human and Language Technology
• /
• 1996.10a
• /
• pp.322-325
• /
• 1996

이 논문은 개념기반의 대화체 언어번역시스템의 번역부의 평가에 대해 기술한 것이다. 대상언어는 한국어와 영어로 한국어를 해석하여 영어로 번역하는 시스템이다. 개념기반 시스템은 개념을 기준으로 입력된 문장을 해석하고 그 개념을 이용하여 번역한다. 개념기반 시스템은 개념에 기반을 두기 때문에 자유로운 간투사의 사용, 빈번한 단어 생략 등의 특성을 가지는 대화체 번역에 유리하다. 시스템의 평가는 입력문에 대한 번역문의 결과가 의미적으로 어느 정도 전달되었는지를 평가자의 주관적인 판단에 의해 평가한다. 현재 개발된 시스템은 여행안내 영역(domain)을 대상으로 하고 있다. 개발된 시스템에 대한 평가는 대화체를 전사한 문장과 음성인식의 결과 두 가지의 입력에 대해 하였다.

• Cross-Cultural Studies
• /
• v.46
• /
• pp.411-434
• /
• 2017

With emphasis on various local cultures to confront the Western central culture, Ngugi wa Thiong'o proposes them 'to move horizontally' so as not to repeat the oppressive culture of the West. We need not only dialogues between dominant languages and peripheral languages, but also between marginal languages. With respect to this point, Ngugi thinks that translation itself could be very effective. Ngugi wants to stimulate writing and speaking in marginalized languages and promote translation as a means of making these languages visible. He regards translation as a conversational tool among languages and cultures in the multicultural global community. As is already well known, his determination to write his later works only in his native Gikuyu language has a great meaning in his anti-colonial as well as anti-neocolonial movement. Its proof is his recent effort to cooperate with Jalada Africa. Simon Gikandi criticized the English translation of Matigari as a denial of cultural hegemony of Gikuyu language and its subordination to the global cultural market. However, the concept of 'thick translation', helps us move from Gikandi's doubt of the 'epistemology of translation' to a meaningful strategy of postcolonial translation. Facing some of the scholars' doubts related to his over-stressing language problem, Ngugi points out that the world has managed to function well through translation: the possibility of translation between cultures and translation as a mediating tool for communication nationally as well as internationally. Based on this two-sided solution of translation, he believes that we can overcome the opposition between relativity and universality, center and periphery, and the dominant and the subordinate.

• Journal of the Korea Convergence Society
• /
• v.12 no.10
• /
• pp.27-35
• /
• 2021

Automatic post editing is a research field that aims to automatically correct errors in machine translation results. This research is mainly being focus on high resource language pairs, such as English-German. Recent APE studies are mainly adopting transfer learning based research, where pre-training language models, or translation models generated through self-supervised learning methodologies are utilized. While translation based APE model shows superior performance in recent researches, as such researches are conducted on the high resource languages, the same perspective cannot be directly applied to the low resource languages. In this work, we apply two transfer learning strategies to Korean-English APE studies and show that transfer learning with translation model can significantly improves APE performance.

• Acta Via Serica
• /
• v.3 no.2
• /
• pp.151-160
• /
• 2018

There is a variety of opinions about the first translation activities within the Turkic Empire. It is widely believed that some Buddhist sutras were translated into the Turkic language in the period of Taspar Qagan (572-581). This theory is based on certain arguments: Some Turks practiced Buddhism, Buddhist monks translated sutras in the center of the Turkic Empire, Taspar brought sutras from China and had them translated, and the monarch of Northern Qi had a sutra translated and sent to Taspar. However, in my opinion, these arguments lack credibility. This article, which is based on primary Chinese sources, will question the likelihood of such translation activities having occurred. Some Chinese records for these claims exist: Da Tang Nei Dian Lu (大唐內典錄) and Xu Gao Seng Chuan (續高僧傳) by the Buddhist monk Jinagupta and the records of Hui Lin in Sui Shu (隋書) and Wen Xian Tong Kao (文獻通考). These are known as "primary sources." Secondary sources, namely contemporary history and language studies, such as those in books and articles, must be based on primary sources. It can be seen that claims relating to the first Turkic translation activities at the time of Taspar are mainly derived from secondary sources, and that the arguments in these secondary sources vary. Sometimes researchers make suppositions on the existence of information that is not referred to in primary sources. However, this is not normal practice. If a researcher relies on unknowns for the evidence of information existing, it can cause false information, ideas and anachronisms to be created. It is important that primary sources, such as the Chinese sources mentioned above, be translated correctly in language and history studies. If only a word is mistranslated, very different results may occur. Mistranslating or misinterpreting a primary source allows conclusions to be reached that are not supported by dissemination of information from primary sources. This can mislead experts and result in information that is not correct being considered as being true. As well as helping to prevent such misinterpretations occurring, another aim of this paper is to question the interpretations of the first Turkic translations in contemporary studies on history and language. The origin of such assessments will be explored and the validity of that information will be examined.

• Journal of Digital Convergence
• /
• v.19 no.7
• /
• pp.199-208
• /
• 2021

Automatic Post Editing(APE) is the study that automatically correcting errors included in the machine translated sentences. The goal of APE task is to generate error correcting models that improve translation quality, regardless of the translation system. For training these models, source sentence, machine translation, and post edit, which is manually edited by human translator, are utilized. Especially in the recent APE research, multilingual pretrained language models are being adopted, prior to the training by APE data. This study deals with multilingual pretrained language models adopted to the latest APE researches, and the specific application method for each APE study. Furthermore, based on the current research trend, we propose future research directions utilizing translation model or mBART model.