• 제목/요약/키워드: Parallel corpus

검색결과 66건 처리시간 0.022초

한영 병렬 코퍼스 구축을 위한 하이브리드 기반 문장 자동 정렬 방법 (A Hybrid Sentence Alignment Method for Building a Korean-English Parallel Corpus)

  • 박정열;차정원
    • 대한음성학회지:말소리
    • /
    • 제68권
    • /
    • pp.95-114
    • /
    • 2008
  • The recent growing popularity of statistical methods in machine translation requires much more large parallel corpora. A Korean-English parallel corpus, however, is not yet enoughly available, little research on this subject is being conducted. In this paper we present a hybrid method of aligning sentences for Korean-English parallel corpora. We use bilingual news wire web pages, reading comprehension materials for English learners, computer-related technical documents and help files of localized software for building a Korean-English parallel corpus. Our hybrid method combines sentence-length based and word-correspondence based methods. We show the results of experimentation and evaluate them. Alignment results from using a full translation model are very encouraging, especially when we apply alignment results to an SMT system: 0.66% for BLEU score and 9.94% for NIST score improvement compared to the previous method.

  • PDF

공공 한영 병렬 말뭉치를 이용한 기계번역 성능 향상 연구 (A Study on the Performance Improvement of Machine Translation Using Public Korean-English Parallel Corpus)

  • 박찬준;임희석
    • 디지털융복합연구
    • /
    • 제18권6호
    • /
    • pp.271-277
    • /
    • 2020
  • 기계번역이란 소스언어를 목적언어로 컴퓨터가 번역하는 소프트웨어를 의미하며 규칙기반, 통계기반 기계번역을 거쳐 최근에는 인공신경망 기반 기계번역에 대한 연구가 활발히 이루어지고 있다. 인공신경망 기계번역에서 중요한 요소 중 하나로 고품질의 병렬 말뭉치를 뽑을 수 있는데 이제까지 한국어 관련 언어쌍의 고품질 병렬 코퍼스를 구하기 쉽지 않은 실정이었다. 최근 한국정보화진흥원의 AI HUB에서 고품질의 160만 문장의 한-영 기계번역 병렬 말뭉치를 공개하였다. 이에 본 논문은 AI HUB에서 공개한 데이터 및 현재까지 가장 많이 쓰인 한-영 병렬 데이터인 OpenSubtitles와 성능 비교를 통해 각각의 데이터의 품질을 검증하고자 한다. 테스트 데이터로 한-영 기계번역 관련 공식 테스트셋인 IWSLT에서 공개한 테스트셋을 이용하여 보다 객관성을 확보하였다. 실험결과 동일한 테스트셋으로 실험한 기존의 한-영 기계번역 관련 논문들보다 좋은 성능을 보임을 알 수 있었으며 이를 통해 고품질 데이터의 중요성을 알 수 있었다.

한국어 연결어미 '-면서'와 중국어 대응표현의 대조연구 -한·중 병렬 말뭉치를 기반으로 (A Comparative Study on Korean Connective Morpheme '-myenseo' to the Chinese expression - based on Korean-Chinese parallel corpus)

  • YI, CHAO
    • 비교문화연구
    • /
    • 제37권
    • /
    • pp.309-334
    • /
    • 2014
  • This study is based on the Korean-Chinese parallel corpus, utilizing the Korean connective morpheme '-myenseo' and contrasting with the Chinese expression. Korean learners often struggle with the use of Korean Connective Morpheme especially when there is a lexical gap between their mother language. '-myenseo' is of the most use Korean Connective Morpheme, it usually contrast to the Chinese coordinating conjunction. But according to the corpus, the contrastive Chinese expression to '-myenseo' is more than coordinating conjunction. So through this study, can help the Chinese Korean language learners learn easier while studying '-myenseo', because the variety Chinese expression are found from the parallel corpus that related to '-myenseo'. In this study, firstly discussed the semantic features and syntactic characteristics of '-myenseo'. The significant semantic features of '-myenseo' are 'simultaneous' and 'conflict'. So in this chapter the study use examples of usage to analyse the specific usage of '-myenseo'. And then this study analyse syntactic characteristics of '-myenseo' through the subject constraint, predicate constraints, temporal constraints, mood constraints, negatives constraints. then summarize them into a table. And the most important part of this study is Chapter 4. In this chapter, it contrasted the Korean connective morpheme '-myenseo' to the Chinese expression by analysing the Korean-Chinese parallel corpus. As a result of the analysis, the frequency of the Chinese expression that contrasted to '-myenseo' is summarized into

    . It can see from the table that the most common Chinese expression comparative to '-myenseo' is non-marker patterns. That means the connection of sentence in Korean can use connective morpheme what is a clarifying linguistic marker, but in Chinese it often connect the sentence by their intrinsic logical relationships. So the conclusion of this chapter is that '-myenseo' can be comparative to Chinese conjunction, expression, non-marker patterns and liberal translation patterns, which are more than Chinese conjunction that discovered before. In the last Chapter, as the conclusion part of this study, it summarized and suggest the limitations and the future research direction.

  • 웹 번역문서 판별과 병렬 말뭉치 구축 (Judging Translated Web Document & Constructing Bilingual Corpus)

    • Jee-hyung, Kim;Yill-byung, Lee
      • 한국정보과학회:학술대회논문집
      • /
      • 한국정보과학회 2004년도 가을 학술발표논문집 Vol.31 No.2 (1)
      • /
      • pp.787-789
      • /
      • 2004
    • People frequently feel the need of a general searching tool that frees from language barrier when they find information through the internet. Therefore, it is necessary to have a multilingual parallel corpus to search with a word that includes a search keyword and has a corresponding word in another language, Multilingual parallel corpus can be built and reused effectively through the several processes which are judgment of the web documents, sentence alignment and word alignment. To build a multilingual parallel corpus, multi-lingual dictionary should be constructed in each language and HTML should be simplified. And by understanding the meaning and the statistics of document structure, judgment on translated web documents will be made and the searched web pages will be aligned in sentence unit.

    • PDF

    정렬된 성경 코퍼스로부터 바꿔쓰기표현(paraphrase)의 자동 추출 (Automatic Extraction of Paraphrases from a Parallel Bible Corpus)

    • 이공주;윤보현
      • 인지과학
      • /
      • 제17권4호
      • /
      • pp.323-336
      • /
      • 2006
    • 바꿔쓰기(paraphrasing)는 동일한 내용을 다르게 표현하는 방식을 의미한다. 이러한 바꿔쓰기표현들(paraphrues)은 기계번역, 질의 응답 시스템, 문서 요약과 같은 다양한 분야에 매우 유용하게 사용될 수 있다. 그러나 이와 같은 바꿔쓰기표현의 유용성에도 불구하고 바꿔쓰기표현을 자동으로 추출할 수 있는 방법이 매우 어렵다. 우선 바꿔쓰기표현을 자동으로 추출할 수 있는 데이터를 구하는 것부터가 어려운 문제이다. 본 연구에서는 여러 버전의 한글 성경 코퍼스로부터 바꿔쓰기표현을 자동으로 추출해 보고자 한다. 성경은 각 문장이 절과 구로 나누어져 있어 문장과 문장을 정렬시키는 것이 매우 용이하다. 정렬된 여러 버전의 성경 코퍼스로부터 자율학습(unsupervised learning)을 통해서 자동으로 바꿔쓰기표현을 추출한다. 이와 같은 방법은 어휘수준의 바꿔쓰기표현 뿐만 아니라 구문수준의 바꿔쓰기표현도 추출할 수 있음을 보여준다.

    • PDF

    한영 병렬 말뭉치를 이용한 한국어 조사 '가'와 '는'의 선택 원리 연구 (The Study on the Principles of Selecting Korean Particle 'Ka' and 'Nun' Using Korean-English Parallel Corpus)

    • 유현경;안예리;양수향
      • 한국언어정보학회지:언어와정보
      • /
      • 제11권1호
      • /
      • pp.1-23
      • /
      • 2007
    • This study aims to research into the meaning of Korean particle 'ka' and 'nun' inductively by examining the correspondences of those particles and English articles on the Korean-English parallel corpus. The correspondences were checked in three ways: semantically, syntactically and pragmatically. This study found that when the semantic or syntactic tier is not salient, the pragmatic tier is activated and particles are selected according to the pragmatic elements such as the amount of information or the change of topic. However, if the meaning of the particles is salient or if there is any syntactic motive, particles are selected in accordance with the semantic or syntactic elements. Former studies which focused on one of those three tiers cannot properly explain such correspondences on the Korean-English parallel corpus. This study shows that semantic, syntactic and pragmatic tiers hierarchically affect the selection of a particle and that the selection process is also related to speaker's intention. This dimensional analysis of particles is expected to contribute to theoretical studies and applied studies like Korean language education as well.

    • PDF

    병렬 말뭉치 필터링을 적용한 Filter-mBART기반 기계번역 연구 (Filter-mBART Based Neural Machine Translation Using Parallel Corpus Filtering)

    • 문현석;박찬준;어수경;박정배;임희석
      • 한국융합학회논문지
      • /
      • 제12권5호
      • /
      • pp.1-7
      • /
      • 2021
    • 최신 기계번역 연구 동향을 살펴보면 대용량의 단일말뭉치를 통해 모델의 사전학습을 거친 후 병렬 말뭉치로 미세조정을 진행한다. 많은 연구에서 사전학습 단계에 이용되는 데이터의 양을 늘리는 추세이나, 기계번역 성능 향상을 위해 반드시 데이터의 양을 늘려야 한다고는 보기 어렵다. 본 연구에서는 병렬 말뭉치 필터링을 활용한 mBART 모델 기반의 실험을 통해, 더 적은 양의 데이터라도 고품질의 데이터라면 더 좋은 기계번역 성능을 낼 수 있음을 보인다. 실험결과 병렬 말뭉치 필터링을 거친 사전학습모델이 그렇지 않은 모델보다 더 좋은 성능을 보였다. 본 실험결과를 통해 데이터의 양보다 데이터의 질을 고려하는 것이 중요함을 보이고, 해당 프로세스를 통해 추후 말뭉치 구축에 있어 하나의 가이드라인으로 활용될 수 있음을 보였다.

    A Corpus Analysis of Temporal Adverbs and Verb Tenses Cooccurrence in Spanish, English, and Chinese

    • Cheng, An Chung;Lu, Hui-Chuan
      • 아시아태평양코퍼스연구
      • /
      • 제3권2호
      • /
      • pp.1-16
      • /
      • 2022
    • This study investigates the cooccurrence between temporal adverbs and grammatical tenses in Spanish and contrasts temporal specifications across Spanish, English, and Chinese. Based on a monolingual Spanish corpus and a trilingual parallel corpus, the study identified the top ten frequent single-word temporal adverbs collocating with grammatical tenses in Spanish. It also contrasted the cooccurrence of temporal adverbs and verb tenses in three languages. The results show that aun 'still', hoy 'today', and ahora 'now' collocate with the present tense at more than 80%. Ayer 'yesterday' and finalmente 'finally' cooccurring with the simple past tense are at 84% and 69%, respectively. Then, mientras 'meanwhile' collocates with the past imperfect at 55%, the highest of all. Mañana 'tomorrow' cooccurs with the future and present tenses at 34%. Other adverbs, ya 'already', siempre 'always', and nuevamete 'again', do not present a strong cooccurrence tendency with a tense overall. The contrastive analysis of the trilingual parallel corpus shows a comprehensive view of temporal specifications in the three languages. However, no clear one-to-one mapping pattern of the cooccurrence across the three languages can be concluded, which provides helpful insights for second language instruction with natural language data rather than intuition. Future research with larger corpora is needed.

    Mining Parallel Text from the Web based on Sentence Alignment

    • Li, Bo;Liu, Juan;Zhu, Huili
      • 한국언어정보학회:학술대회논문집
      • /
      • 한국언어정보학회 2007년도 정기학술대회
      • /
      • pp.285-292
      • /
      • 2007
    • The parallel corpus is an important resource in the research field of data-driven natural language processing, but there are only a few parallel corpora publicly available nowadays, mostly due to the high labor force needed to construct this kind of resource. A novel strategy is brought out to automatically fetch parallel text from the web in this paper, which may help to solve the problem of the lack of parallel corpora with high quality. The system we develop first downloads the web pages from certain hosts. Then candidate parallel page pairs are prepared from the page set based on the outer features of the web pages. The candidate page pairs are evaluated in the last step in which the sentences in the candidate web page pairs are extracted and aligned first, and then the similarity of the two web pages is evaluate based on the similarities of the aligned sentences. The experiments towards a multilingual web site show the satisfactory performance of the system.

    • PDF

    영한 병렬 코퍼스에 나타난 영어 수동문의 한국어 번역 (Translating English By-Phrase Passives into Korean: A Parallel Corpus Analysis)

    • 이승아
      • 영어영문학
      • /
      • 제56권5호
      • /
      • pp.871-905
      • /
      • 2010
    • This paper is motivated by Watanabe's (2001) observation that English byphrase passives are sometimes translated into Japanese object topicalization constructions. That is, the original English sentence in the passive may be translated into the active voice with the logical object topicalized. A number of scholars, including Chomsky (1981) and Baker (1992), have remarked that languages have various ways to avoid focusing on the logical subject. The aim of the present study is to examine the translation equivalents of the English by-phrase passives in an English-Korean parallel corpus compiled by the author. A small sample of articles from Newsweek magazine and its published Korean translation reveals that there are indeed many ways to translate English by-phrase passives, including object topicalization (12.5%). Among the 64 translated sentences analyzed and classified, 12 (18.8%) examples were problematic in terms of agent defocusing, which is the primary function of passives. Of these 12 instances, five cases were identified where an alternative translation would be more suitable. The results suggest that the functional characteristics of English by-phrase passives should be highlighted in translator training as well as language teaching.