• Journal of Multimedia Information System
• /
• v.9 no.2
• /
• pp.93-102
• /
• 2022

Dependency parsing is a decision problem of the syntactic relation between words in a sentence. Recently, deep learning models are used for dependency parsing based on the word representations in a continuous vector space. However, it causes a mislabeled tagging problem for the proper nouns that rarely appear in the training corpus because it is difficult to express out-of-vocabulary (OOV) words in a continuous vector space. To solve the OOV problem in dependency parsing, we explored the proper noun embedding method according to the embedding unit. Before representing words in a continuous vector space, we replace the proper nouns with a special token and train them for the contextual features by using the multi-layer bidirectional LSTM. Two models of the syllable-based and morpheme-based unit are proposed for proper noun embedding and the performance of the dependency parsing is more improved in the ensemble model than each syllable and morpheme embedding model. The experimental results showed that our ensemble model improved 1.69%p in UAS and 2.17%p in LAS than the same arc-eager approach-based Malt parser.

• Smart Media Journal
• /
• v.8 no.2
• /
• pp.74-79
• /
• 2019

Traditional compound noun decomposition algorithms often face challenges of decomposing compound nouns into separated nouns when unregistered unit noun is included. It is very difficult for those traditional approach to handle such issues because it is impossible to register all existing unit nouns into the dictionary such as proper nouns, coined words, and foreign words in advance. In this paper, in order to solve this problem, compound noun decomposition problem is defined as tag sequence labeling problem and compound noun decomposition method to use syllable unit embedding and deep learning technique is proposed. To recognize unregistered unit nouns without constructing unit noun dictionary, compound nouns are decomposed into unit nouns by using LSTM and linear-chain CRF expressing each syllable that constitutes a compound noun in the continuous vector space.

• Journal of KIISE
• /
• v.44 no.3
• /
• pp.306-313
• /
• 2017

Named entity recognition (NER) seeks to locate and classify named entities in text into pre-defined categories such as names of persons, organizations, locations, expressions of times, etc. Recently, many state-of-the-art NER systems have been implemented with bidirectional LSTM CRFs. Deep learning models based on long short-term memory (LSTM) generally depend on word representations as input. In this paper, we propose an approach to expand word representation by using pre-trained word embedding, part of speech (POS) tag embedding, syllable embedding and named entity dictionary feature vectors. Our experiments show that the proposed approach creates useful word representations as an input of bidirectional LSTM CRFs. Our final presentation shows its efficacy to be 8.05%p higher than baseline NERs with only the pre-trained word embedding vector.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.17 no.4
• /
• pp.199-208
• /
• 2022

Korean is an agglutinative language, and one or more morphemes are combined to form a single word. Part-of-speech tagging method separates each morpheme from a word and attaches a part-of-speech tag. In this study, we propose a new Korean part-of-speech tagging method based on the Head-Tail tokenization technique that divides a word into a lexical morpheme part and a grammatical morpheme part without decomposing compound words. In this method, the Head-Tail is divided by the syllable boundary without restoring irregular deformation or abbreviated syllables. Korean part-of-speech tagger was implemented using the Head-Tail tokenization and deep learning technique. In order to solve the problem that a large number of complex tags are generated due to the segmented tags and the tagging accuracy is low, we reduced the number of tags to a complex tag composed of large classification tags, and as a result, we improved the tagging accuracy. The performance of the Head-Tail part-of-speech tagger was experimented by using BERT, syllable bigram, and subword bigram embedding, and both syllable bigram and subword bigram embedding showed improvement in performance compared to general BERT. Part-of-speech tagging was performed by integrating the Head-Tail tokenization model and the simplified part-of-speech tagging model, achieving 98.99% word unit accuracy and 99.08% token unit accuracy. As a result of the experiment, it was found that the performance of part-of-speech tagging improved when the maximum token length was limited to twice the number of words.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.1
• /
• pp.305-318
• /
• 2022

Abusive behaviors have become a common issue in many online social media platforms. Profanity is common form of abusive behavior in online. Social media platforms operate the filtering system using popular profanity words lists, but this method has drawbacks that it can be bypassed using an altered form and it can detect normal sentences as profanity. Especially in Korean language, the syllable is composed of graphemes and words are composed of multiple syllables, it can be decomposed into graphemes without impairing the transmission of meaning, and the form of a profane word can be seen as a different meaning in a sentence. This work focuses on the problem of filtering system mis-detecting normal phrases with profane phrases. For that, we proposed the deep learning-based framework including grapheme and syllable separation-based word embedding and appropriate CNN structure. The proposed model was evaluated on the chatting contents from the one of the famous online games in South Korea and generated 90.4% accuracy.

• KIPS Transactions on Software and Data Engineering
• /
• v.7 no.4
• /
• pp.121-128
• /
• 2018

A Korean morphological analyzer adopts sequence-to-sequence (seq2seq) model, which can generate an output sequence of different length from an input. In general, a seq2seq based Korean morphological analyzer takes a syllable-unit based sequence as an input, and output a syllable-unit based sequence. Syllable-based morphological analysis has the advantage that unknown words can be easily handled, but has the disadvantages that morpheme-based information is ignored. In this paper, we propose a reranking model as a post-processor of seq2seq model that can improve the accuracy of morphological analysis. The seq2seq based morphological analyzer can generate K results by using a beam-search method. The reranking model exploits morpheme-unit embedding information as well as n-gram of morphemes in order to reorder K results. The experimental results show that the reranking model can improve 1.17% F1 score comparing with the original seq2seq model.

• Annual Conference on Human and Language Technology
• /
• 2018.10a
• /
• pp.574-579
• /
• 2018

본 논문에서는 한국어 형태소 분석 시스템을 제안하는데, 연구 목표는 오타 없는 문서를 대상으로 한 경우에도 높은 성능을 유지하면서, 동시에 오타가 있는 문서에서도 우수한 성능을 산출하는 것이다. 실험은 크게 두 종류로 나누어서 진행된다. 주 실험인 첫 번째 실험에서는, 자모 임베딩과 음절 임베딩을 결합(concatenate)한 벡터를 입력으로 Bidirectional LSTM CRFs을 수행함으로써, 세종말뭉치 대상으로 어절 정확도 97%, 그리고 1, 2, 5 어절마다 오타가 출현한 경우에서도 각각 80.09%, 87.53%, 92.49%의 높은 성능을 산출하였다. 추가 실험인 두 번째 실험에서는, 실생활에서 자주 발생하는 오타들을 집계하여 그 중에서 11가지 오타 유형을 선정 후, 각 유형에 대해 변환된 임베딩 벡터를 적용함으로써, 해당 오타를 포함한 문장에서 93.05%의 우수한 성능을 산출하였다.

• Annual Conference on Human and Language Technology
• /
• 2018.10a
• /
• pp.605-607
• /
• 2018

본 논문에서는 음절 임베딩과 양방향 LSTM-CRF 모델을 이용한 한국어 문장 자동 띄어쓰기 시스템을 제안한다. 문장에 대한 자질 벡터 표현을 위해 문장을 구성하는 음절을 Unigram 및 Bigram으로 나누어 각 음절을 연속적인 벡터 공간에 표현하고, 양방향 LSTM을 이용하여 현재 자질에 양방향 자질들과 의존성을 부여한 새로운 자질 벡터를 생성한다. 이 새로운 자질 벡터는 전방향 신경망과 선형체인(Linear-Chain) CRF를 이용하여 최적의 띄어쓰기 태그 열을 예측하고, 생성된 띄어쓰기 태그를 기반으로 문장 자동 띄어쓰기를 수행하였다. 문장 13,500개와 277,718개 어절로 이루어진 학습 데이터 집합과 문장 1,500개와 31,107개 어절로 이루어진 테스트 집합의 학습 및 평가 결과는 97.337%의 음절 띄어쓰기 태그 분류 정확도를 보였다.

• Journal of Information Processing Systems
• /
• v.15 no.5
• /
• pp.1119-1130
• /
• 2019

With the wide spread of Social Network Services (SNS), fake news-which is a way of disguising false information as legitimate media-has become a big social issue. This paper proposes a deep learning architecture for detecting fake news that is written in Korean. Previous works proposed appropriate fake news detection models for English, but Korean has two issues that cannot apply existing models: Korean can be expressed in shorter sentences than English even with the same meaning; therefore, it is difficult to operate a deep neural network because of the feature scarcity for deep learning. Difficulty in semantic analysis due to morpheme ambiguity. We worked to resolve these issues by implementing a system using various convolutional neural network-based deep learning architectures and "Fasttext" which is a word-embedding model learned by syllable unit. After training and testing its implementation, we could achieve meaningful accuracy for classification of the body and context discrepancies, but the accuracy was low for classification of the headline and body discrepancies.

• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.11
• /
• pp.441-448
• /
• 2019

Previous researches on automatic spacing of Korean sentences has been researched to correct spacing errors by using n-gram based statistical techniques or morpheme analyzer to insert blanks in the word boundary. In this paper, we propose an end-to-end automatic word spacing by using deep neural network. Automatic word spacing problem could be defined as a tag classification problem in unit of syllable other than word. For contextual representation between syllables, Bi-LSTM encodes the dependency relationship between syllables into a fixed-length vector of continuous vector space using forward and backward LSTM cell. In order to conduct automatic word spacing of Korean sentences, after a fixed-length contextual vector by Bi-LSTM is classified into auto-spacing tag(B or I), the blank is inserted in the front of B tag. For tag classification method, we compose three types of classification neural networks. One is feedforward neural network, another is neural network language model and the other is linear-chain CRF. To compare our models, we measure the performance of automatic word spacing depending on the three of classification networks. linear-chain CRF of them used as classification neural network shows better performance than other models. We used KCC150 corpus as a training and testing data.