• 대한전기학회논문지:시스템및제어부문D
• /
• 제51권10호
• /
• pp.446-452
• /
• 2002

For most of pattern recognition applications, it is required to correctly recognize patterns even if they have translation variations. In this paper, to achieve the goal of translation invariant pattern recognition, we propose a new generalized translation invariant second-order neural network using a constraint on the weights. The weight constraint is implemented using generalized translation invariant features which are accumulated sums of pixel combinations. Simulation results will be given to demonstrate that the proposed second-order neural network has the generalized translation invariant property.

• 한국언어정보학회지:언어와정보
• /
• 제2권1호
• /
• pp.1-41
• /
• 1998

Evaluation of multilingual translation fundamentally involves measurement of meaning equivalences between the formally mapped discourses/texts of SL(source language) and TL(target language) both represented by a metalanguage called IL(interlingua). Unlike a usaal uni-directional MT(machine translation) model(e.g.:SL $\rightarrow$ analysis $\rightarrow$ transfer $\rightarrow$ generation $\rightarrow$ TL), a bi-directional(by 'negotiation') model(i.e.: SL $\rightarrow$ IL/S $\leftrightarrow$ IL $\leftrightarrow$ IL/T \leftarrow TL) is proposed here for the purpose of evaluating multilingual, not merely bilingual, translation. The IL, as conceived of in this study, is an English-based predicate logic represented in the framework of MRS(minimal recursion semantics), an MT-oriented off-shoot of HPSG(Head-driven Phrase Structure Grammar). In addition, a list of semantic and pragmatic checkpoints are set up, some being optional depending on the kind and use of the translation, so sa to have the evaluation of translation fine-grained by computing matching or mismatching of such checkpoints.

• Communications for Statistical Applications and Methods
• /
• 제27권3호
• /
• pp.313-326
• /
• 2020

Deep Learning is the most important key to the development of Artificial Intelligence (AI). There are several distinguishable architectures of neural networks such as MLP, CNN, and RNN. Among them, we try to understand one of the main architectures called Recurrent Neural Network (RNN) that differs from other networks in handling sequential data, including time series and texts. As one of the main tasks recently in Natural Language Processing (NLP), we consider Neural Machine Translation (NMT) using RNNs. We also summarize fundamental structures of the recurrent networks, and some topics of representing natural words to reasonable numeric vectors. We organize topics to understand estimation procedures from representing input source sequences to predict target translated sequences. In addition, we apply multiple translation models with Gated Recurrent Unites (GRUs) in Keras on English-Korean sentences that contain about 26,000 pairwise sequences in total from two different corpora, colloquialism and news. We verified some crucial factors that influence the quality of training. We found that loss decreases with more recurrent dimensions and using bidirectional RNN in the encoder when dealing with short sequences. We also computed BLEU scores which are the main measures of the translation performance, and compared them with the score from Google Translate using the same test sentences. We sum up some difficulties when training a proper translation model as well as dealing with Korean language. The use of Keras in Python for overall tasks from processing raw texts to evaluating the translation model also allows us to include some useful functions and vocabulary libraries as well.

• 한국어병학회지
• /
• 제34권1호
• /
• pp.17-22
• /
• 2021

Eukaryotic translation is initiated by either cap-dependent or cap-independent way, and the cap-independent translation can be initiated by the internal ribosomal entry site (IRES). In this study, to know whether the 5'UTR leader sequence of marine birnavirus (MABV) segment A and segment B can act as IRES, bicistronic vectors harboring a CMV promoter-driven red fluorescent gene (mCherry) and poliovirus IRES- or MABV's leader sequence-driven green fluorescent gene (eGFP) were constructed, then, transfected into a mammalian cell line (BHK-21 cells) and a fish cell line (CHSE-214 cells). The results showed that the poliovirus IRES worked well in BHK-21 cells, but did not work in CHSE-214 cells. In the evaluation of MABV's leader sequences, the reporter eGFP gene under the 5'UTR leader sequence of MABV's segment A was well-translated in CHSE-214 cells, indicating 5'UTR of MABV's segment A initiates translation in the cap-independent way and can be used as a fish-specific IRES system. However, the 5'UTR leader sequence of MABV's segment B did not initiate translation in CHSE-214 cells. As the precise mechanism of birnavirid IRES-mediated translation is not known, more elaborate investigations are needed to uncover why the leader sequence of segment B could not initiate translation in the present study. In addition, further studies on the host species range of MABV's segment A IRES and on the screening of other fish-specific IRESs are needed.

• 한국정보과학회논문지:소프트웨어및응용
• /
• 제41권7호
• /
• pp.501-517
• /
• 2014

이진 변환은 특정 장치에서 동작하도록 구성된 프로그램을 다른 장치에서 동작할 수 있도록 재구성하는 과정을 말한다. 이진 변환을 수행하기 위해서는 두 장치 사이의 변환 규칙을 생성하는 것이 매우 중요하다. 변환 규칙을 생성하는 방법은 직접 매핑과 간접 매핑으로 나뉜다. 직접 매핑은 성능을 위한 방법인 반면, 간접 매핑은 재목적성을 위한 방법이다. 본 논문에서는 임베디드 시스템에 적합한 직접 매핑 기반의 이진 변환을 수행한다. 그렇지만 재목적성 역시 중요한 요구사항이기 때문에, 재목적성을 고려한 직접 매핑 기반의 이진 변환 방법을 제안한다. 또한 제안된 방법을 바탕으로 자동으로 변환 규칙을 생성하는 도구를 구현한다. 이 방법을 통해서 성능과 재목적성을 모두 고려한 변환 규칙을 생성할 수 있으며, 더 나아가 이진 변환을 수행하는데 소요되는 비용을 줄일 수 있다.

• International Journal of Advanced Culture Technology
• /
• 제9권3호
• /
• pp.259-264
• /
• 2021

In modern socity, online communication plays a vital role in social interaction of communicities. It is so common for online users to see display advertisements online while surting the Net. Specifically, most web banners diaplayed on portral sites consist of words, phrase, and sentences. Considering that the primary purpose of adversiting is persuation, the advertisement such as web banners is an examplary case to show the interaction among pragmatics, translation and advertising because the linguistic expressions employed in the banners represent its pragmatic use, leading to persuation and functioning as a communicative tool for the smooth communication between source text producers (adversisers) and target audience (online users). This can be part of the so-called translation process. In particular, we can easily witness the use of English words in web banners. Thus, this paper looks at web banners displayed on major four portal sites-Naver, Daum, Nate, and Zum, giving a special attention to the content contained in the web banners as well as the use of English words. As s result, we found that the frequencies of English words in each portal site were higher when the advertised products were targeting young online users, whereas the frequencies were lower when the users are older group than young people. The finding supports the prgramatic perspective that linguistic expressions are understood in social contexts and shows the so-called translation process which involves a shift from semantic meaning of words to their pragmatic use. Finally, we can conclude that the interaction is possible when we have the framework where translation, pragmatics, and advertising are all communitative components for social interaction within social contexts.

• 디지털융복합연구
• /
• 제19권3호
• /
• pp.25-35
• /
• 2021

인터넷과 연결 가능한 기기만 있다면 어디서든 외국의 영상매체를 시청할 수 있게 되었다. Netflix와 Youtube는 대표적인 영상 플랫폼이고 많은 사람들이 이용하고 있기도 하다. 그러다보니 외국어를 알지 못하는 경우, 전적으로 자막에 의존할 수밖에 없다. 이에 본 연구에서는 두 개의 인기 플랫폼에서 양질의 번역을 제공하고 있는지 알아보기 위해 두 곳에서 동시에 상영하고 있는 영화를 골라 한국어 번역본을 가지고 비교해보고 더 나은 방향과 문제점을 도출하였다. 그 결과, 좋은 번역이란 번역가가 시청자들이 집중을 할 수 있게 적절한 생략과 자세한 설명을 한정된 시간과 공간 안에 할 수 있어야 한다. 작업물을 심도 있게 연구하고 충분한 시간을 들여 작업에 임한다면 번역의 퀼리티는 확연히 올라갈 것이다. 마지막으로 본 연구는 영상번역에 대한 가이드라인 없이 싼 단가에 최소한의 시간을 들여 제작한 영상물들의 오역을 통해 아직 과거에 머물고 있는 영상번역 시장 활성화에 기여하였다.

• 대한한의학원전학회지
• /
• 제24권1호
• /
• pp.57-71
• /
• 2011

Although there have been many endeavors aimed at the standardization and globalization of Korean medicine over a long period of time, the access to information on Oriental medical classics has been relatively poor due to the lack of appropriate translation methodology and standard terminology. In order to overcome existing barriers, continuous effort towards precise translation adopting a standard terminology should be maintained. As a part of this procedure, we planned to publish a part of "Classic of difficult issues(難經)" in three sections, and the first two studies have already been published. Based on the methodology and approaches of previous studies, this third study aims to translate parts of "Classic of difficult issues(難經)" into English, beginning with "The Sixteenth Question", and adopting "WHO-IST" terminology. The outcomes of this study are presented as follows: First, based on the result of existing translation studies and the outcome of "WHO-IST", English translation of "Classic of difficult issues(難經)" from "the Sixteenth Question" to "The Twenty-second Question" is offered, hoping to set a model of translation study which can be communicated universally. Second, in order to pave the way for future success in establishing translation studies, it is natural to verify the effectiveness and practicality of standard terminologies including the outcome of "WHO-IST". Continuous translation studies will be required in order to obtain constant feedback and adopt more suitable guidelines during the standardization process. Taking this into consideration, further translation studies of Oriental medical classics including "Classic of difficult issues(難經)" should be continued.

• 한국멀티미디어학회논문지
• /
• 제7권7호
• /
• pp.998-1008
• /
• 2004

일본어와 한국어는 문법적으로 많은 유사점을 가지고 있다. 이러한 유사점을 잘 이용한다면 일한 기계번역 시스템에서 구문해석이나 의미해석의 상당한 부분을 생략할 수 있다. 몇 년 전부터 우리는 유사성을 이용하여 번역율을 높이는 방법으로 번역테이블을 이용한 일한기계번역 시스템을 연구해 왔다. 그러나 이 시스템은 활용어미의 번역, 다의성 단어의 처리 등 몇 가지 문제점을 가지고 있었다. 본 논문에서는 번역테이블을 이용하는 시스템을 개선하여 이웃 하는 단어들과의 관계 정보를 이용한 일한 기계번역 시스템을 제안한다. 현재 시스템의 문제점들을 해결하기 위하여 우선 조사, 조동사의 접속 정보를 최대한 이용한다. 또한, 번역 테이블을 엔트리테이블과 접속정보 테이블로 나누어 설계하여 번역의 효율을 높인다. 즉, 하나의 역어만 가지는 단어인 경우, 우리는 일한 직접 대응 방법을 이용하여 바로 번역하고 2개 이상의 역어로 번역되어야 할 경우만 접속 정보 값을 평가하여 가장 가능성이 높은 번역어를 선택하도록 한다.

• 대한한의학회지
• /
• 제31권1호
• /
• pp.57-68
• /
• 2010

Objectives: Standard is a unified criterion for some repeated things or concepts in a certain scope. It is fundamental to implement standardization in English translation of traditional Korean medicine (TKM) formulae to promote progress in the evaluation of TKM and also to serve in enhancing the efficiency in studies of medical formulae. Methods: We undertook literature research on current Korean and Chinese medicinal formulae in English translation, analyzing 485 Korean formulae and 464 Chinese. We also undertook a comparative study of 102 common English translation of both Korean and Chinese, proposing a constant and effective methods for English translation of medical formulae. Results: To have a precise English translation we classified medical formulae nomenclature in advance. We found that formulae naming can be fundamentally classified into 6 forms which are (a) Materia Medica + Preparation Form, (b) Materia Medica + Indication + Preparation Form, (c) Materia Medica Numbers + Preparation Form, (d) Indication + Preparation Form, (e) Concept + Preparation Form, and (f) Miscellaneous. Based on these findings we could determine that these 6 types ((a)-(f)) can all be translated into English by (1) Materia Medica + Preparation Form and (2) Indication + Preparation Form + of (with) + Materia Medica schemes. In regard to translation of Indication it can follow Noun+~ing participle form. Conclusion: This research provides a common method of TKM medicinal formulae English translation for better understanding, education, training, practice and research in TKM. Consequently, English translation using these methods can serve as the initial study for standardization of TKM medicinal formulae.