• International Journal of Advanced Culture Technology
• /
• v.7 no.2
• /
• pp.179-187
• /
• 2019

As the globalization process progresses more rapidly and widely, there has been a ever-growing demand for multilingual learning. Compared with the study of Second Language Acquisition, studies on the Third Language and even Multilingual Acquisition have attracted a relatively poor attention. At the same time, considering current educational environments for ethnic colleges and universities, the effects of college English teaching for minority students can be said to have been 'generally poor.' In this situation, when we try to find ways to improve college English studies for minority students most of whom already can speak two languages or even more, it would not be the best idea to keep following the experiences of traditional Second Language Acquisition. It is necessary first to find out whether there are positive or negative effects in acquiring multiple languages, and then to conduct a profound research on L3 (third language and even multilingual) Acquisition in order to employ more efficient teaching methods for multilingual learners. After conducting a Japanese-teaching experiment on two groups of learners with mono-lingual and bilingual backgrounds, it has been found that there is a positive transfer between different languages. In this paper, following the recent research findings on Language teaching for multilingual learners, I try to show with further supports that when it comes to language education for learners with multilingual backgrounds, we should focus on the advantages they may earn in order to conduct more effective language acquisition.

• Journal of Information Processing Systems
• /
• v.18 no.4
• /
• pp.535-548
• /
• 2022

Using a rich resource language to classify sentiments in a language with few resources is a popular subject of research in natural language processing. Burmese is a low-resource language. In light of the scarcity of labeled training data for sentiment classification in Burmese, in this study, we propose a method of transfer learning for sentiment analysis of a language that uses the feature transfer technique on sentiments in English. This method generates a cross-language word-embedding representation of Burmese vocabulary to map Burmese text to the semantic space of English text. A model to classify sentiments in English is then pre-trained using a convolutional neural network and an attention mechanism, where the network shares the model for sentiment analysis of English. The parameters of the network layer are used to learn the cross-language features of the sentiments, which are then transferred to the model to classify sentiments in Burmese. Finally, the model was tuned using the labeled Burmese data. The results of the experiments show that the proposed method can significantly improve the classification of sentiments in Burmese compared to a model trained using only a Burmese corpus.

• ETRI Journal
• /
• v.46 no.1
• /
• pp.59-70
• /
• 2024

We introduce a high-performance named entity recognition (NER) model for written and spoken language. To overcome challenges related to labeled data scarcity and domain shifts, we use transfer learning to leverage our previously developed KorBERT as the base model. We also adopt a meta-pseudo-label method using a teacher/student framework with labeled and unlabeled data. Our model presents two modifications. First, the student model is updated with an average loss from both human- and pseudo-labeled data. Second, the influence of noisy pseudo-labeled data is mitigated by considering feedback scores and updating the teacher model only when below a threshold (0.0005). We achieve the target NER performance in the spoken language domain and improve that in the written language domain by proposing a straightforward rollback method that reverts to the best model based on scarce human-labeled data. Further improvement is achieved by adjusting the label vector weights in the named entity dictionary.

• Language and Information
• /
• v.18 no.1
• /
• pp.77-97
• /
• 2014

Multilingual machine translation means the machine translation that is for more than two languages. Common transfer means the transfer in which we can reuse the transfer rules among similar languages according to linguistic typology. Therefore, the multilingual machine translation based on common transfer is the multilingual machine translation that can share the transfer rules among languages with similar linguistic typology. This paper describes the linguistic modeling for multilingual machine translation based on common transfer under development. This linguistic modeling consists of the linguistic devices such as 1) multilingual common Part-of-Speech set, 2) multilingual common transfer format, 3) multilingual common transfer chunking, and 4) multilingual common transfer rules based on linguistic typology. Validity of this linguistic modeling for multilingual machine translation is shown in the simulation. The multilingual machine translation system based on common transfer including Korean, English, Chinese, Spanish, and French will be developed till 2018.

• Speech Sciences
• /
• v.15 no.1
• /
• pp.25-36
• /
• 2008

Language transfer means that the speaker's first language or previously acquired language influences on the production of the target language. This study aims at examining if there is native language influence on the production of English intonation by Korean speakers. The pitch accent patterns and the values of duration, F0, and intensity of the stressed vowel of the word with emphatic accent in the sentence produced by Korean speakers are compared to those of American English speakers. The results show that when the word receives emphatic accent in the sentence, American English speakers put H* accent on the stressed syllable of the word, but Korean speakers mostly assign high pitch on the last syllable of the word and have LH tonal pattern despite the fact that primary stress does not come on the last syllable within a word. In addition, comparison of the values of duration, F0, and intensity of the stressed vowel of the word with emphatic accent to those of the word with unmarked neutral accent shows that Korean speakers do not realize the intonation of the accented word appropriately because the values decrease even though the word has emphatic accent. This study finds out that there are differences in the production of English intonation of the word with emphatic accent between native speakers of English and Korean speakers, and that there is negative transfer of Korean intonation pattern to the production of English intonation by Korean speakers.

• Phonetics and Speech Sciences
• /
• v.3 no.3
• /
• pp.77-88
• /
• 2011

Foreign accent in second language production depends heavily on the transfer of features from the first language. This study examines acoustic variations in segments and suprasegments by native and nonnative speakers of English, searching for patterns of the transfer and plausible indexes of foreign accent in English. The acoustic variations are analyzed with recorded read speech by 20 native English speakers and 50 Korean learners of English, in terms of vowel formants, vowel duration, and syllabic variation induced by stress. The results show that the acoustic measurements of vowel formants and vowel and syllable durations display difference between native speakers and nonnative speakers. The difference is robust in the production of lax vowels, diphthongs, and stressed syllables, namely the English-specific features. L1 transfer on L2 specification is found both at the segmental levels and at the suprasegmental levels. The transfer levels measured as groups and individuals further show a continuum of divergence from the native-like target. Overall, the eldest group, students who are in the graduate schools, shows more native-like patterns, suggesting weaker foreign accent in English, whereas the high school students tend to involve larger deviation from the native speakers' patterns. Individual results show interdependence between segmental transfer and prosodic transfer, and correlation with self-reported proficiency levels. Additionally, experience factors in English such as length of English study and length of residence in English speaking countries are further discussed as factors to explain the acoustic variation.

• Phonetics and Speech Sciences
• /
• v.2 no.3
• /
• pp.29-36
• /
• 2010

The purposes of this study are (1) to see if language transfer effect is found in Korean speakers' pronunciation of English stops and to correct them and (2) to investigate the effectiveness of mimicry training and Speech Analyzer training on subjects' pronunciation of English stops. For these purposes, 20 Korean speakers' VOT values of English stops were measured using Speech Analyzer and their post-training production was compared with their pre-training production. The result shows that Korean speakers have no difficulty in correcting pronunciation errors of English voiceless stops and voiced stops and such a result indicates that language transfer effect is not noticed as expected. In addition, the result of pronunciation training shows that the training using Speech Analyzer is more effective than mimicry training.

• Journal of the Korea Convergence Society
• /
• v.12 no.11
• /
• pp.35-43
• /
• 2021

Recently, there has been a growing interest in zero-shot cross-lingual transfer, which leverages cross-lingual language models (CLLMs) to perform downstream tasks that are not trained in a specific language. In this paper, we point out the limitations of the data-centric aspect of quality estimation (QE), and perform zero-shot cross-lingual transfer even in environments where it is difficult to construct QE data. Few studies have dealt with zero-shots in QE, and after fine-tuning the English-German QE dataset, we perform zero-shot transfer leveraging CLLMs. We conduct comparative analysis between various CLLMs. We also perform zero-shot transfer on language pairs with different sized resources and analyze results based on the linguistic characteristics of each language. Experimental results showed the highest performance in multilingual BART and multillingual BERT, and we induced QE to be performed even when QE learning for a specific language pair was not performed at all.

• Annual Conference on Human and Language Technology
• /
• 1996.10a
• /
• pp.301-307
• /
• 1996

다국어 자동번역이란 2개국어 이상 언어들간의 번역을 말한다. 기존의 다국어 자동번역 시스템은 크게 변환기반 transfer-based 방식과 피봇방식으로 분류될 수 있는데 변환기반 다국어 자동번역 시스템에서는 각 언어의 분석과 생성 규칙이 상이하게 작성됨으로써 언어들간의 공통성이 수용되지 못하였고 그로 인해 전체 번역 메모리의 크기가 증가하는 결과를 초래하였었다. 또한 기존의 피봇방식에서는 다국어에 적용될 수 있는 언어학적 보편성 모델을 구현하는 어려움이 있었다. 이러한 기존의 다국어 자동번역 시스템의 단점들을 극복하기 위해 본 논문에서는 언어들간의 공통성을 수용하며 또한 여러 언어에서 공유될 수 있는 공통 규칙에 의한 다국어 자동번역 시스템을 제안하고자 한다. 공통 규칙의 장점은 전산학적으로는 여러 언어에서 단지 한번 load 되기 때문에 전체 번역 메모리의 크기를 줄일 수 있다는 것과 언어학적으로는 문법 정보의 작성.수정.관리의 일관성을 유지할 수 있다는 것이다.

• English Language & Literature Teaching
• /
• no.1
• /
• pp.159-187
• /
• 1995

The purpose of this study is to investigate common errors made by Korean learners in English composition and to find out what is an effective and appropriate teaching method of English composition in Korea. For these purposes, 197 students on the third grade in high school were selected as the subjects of this research. The students were tested by way of the immediate translation of 31 simple Korean sentences into English which are supposed to be easy for those students to write without any difficulty. About 2 minutes were given for testing each sentence. The results are as follows : First. the whole sum of errors made by 197 students was 2,972 and these types of errors were classified into 13 categories by Duskova's grammatical method and James'. The errors with comparatively high frequency were prepositional errors(17.2%), verbal errors(15.4%), and the errors with low frequency were article errors(1.9%), to-infinitive errors. Second, when Korean students learn English as a target language, overgeneralization(33.6%) and reduction(17.5) influenced the learners much more greatly than language transfer(22.2) did. But the influence of language transfer including interference & overgeneralization(l5.2%) and interference & reduction(10.7%) was no less than 48.1%. The statistics shows that the learners have a tendency to analyze, systematize and regularize the target language when they start to learn a new language.