• The Journal of the Acoustical Society of Korea
• /
• v.20 no.2E
• /
• pp.3-8
• /
• 2001

In this paper, we will examine a Korean large vocabulary continuous speech recognition (LVCSR) system for broadcast news speech. The combined vowel and implosive unit is included in a phone set together with other short phone units in order to obtain a longer unit acoustic model. The effect of this unit is compared with conventional phone units. The dictionary units for language processing are automatically extracted from eojeols appearing in transcriptions. Triphone models are used for acoustic modeling and a trigram model is used for language modeling. Among three major speaker groups in news broadcasts-anchors, journalists and people (those other than anchors or journalists, who are being interviewed), the speech of anchors and journalists, which has a lot of noise, was used for testing and recognition.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1998.08a
• /
• pp.74-77
• /
• 1998

우리는 C++를 이용하여 음성인식기를 구현하여 기존의 C를 이용한 경우에 비하여 30% 수준의 소스로 표현하였고 인식기의 공동개발, 확장 및 개선, 기술 전수 등이 용이하게 되었으며 이를 음성인식 엔진 및 음성인식 연구를 위한 툴로 사용할 수 있게 되었다. 이 인식기의 특징으로는 연속 음성 및 대화체 음성을 인식할 수 있으며 trigram 언어 모델을 사용하였고 문맥 종속 음소 모델링에서는 기존의 triphone 보다 넓은 문맥을 고려한 n-phone context modeling을 사용하였으며 모델의 선정에는 음성학적 지식을 기반으로 한 질문을 사용한 decision tree를 사용하여 훈련에 나타나지 않은 단어나 문맥인 경우라도 가장 가까운 모델을 선정할 수 있게 하였다. 또, tree lexicon을 사용하여 속도를 개선하였으며 state 단위의 모델 공유를 통해 제한된 데이터를 이용하여 더 많은 모델을 훈련할 수 있어 성능을 개선하였다. 상용화를 염두에 두고 pc에서 구현하였다.

• Annual Conference on Human and Language Technology
• /
• 1996.10a
• /
• pp.203-206
• /
• 1996

본 논문에서는 음소단위의 bigram과 trigram 정보를 이용하여 어절내에서의 위치와 개수에 관계없이 미등록어를 추정하고, 미등록어용 형태소 패턴 사전을 도입하여 마치 등록어처럼 미등록어를 처리할 수 있는 방법을 제안한다. 제안된 미등록어 추정 모텔은 조사나 어미와 같은 기능어에 의한 간접적인 추정방법이 아닌 미등록어 자체의 추정과 접속정보를 이용한 검사를 동시에 하여 정확도를 높였다. 본 미등록어 추정방법은 기존의 한국어 품사태깅모델인 TAKTAG에 적용하여 미등록어가 포함된 어절에 대해서 83.72%의 성능을 보였다.

• Phonetics and Speech Sciences
• /
• v.6 no.3
• /
• pp.155-164
• /
• 2014

This paper proposes a new method to determine the recognition units for large vocabulary continuous speech recognition (LVCSR) in Korean by applying unsupervised segmentation and merging. In the proposed method, a text sentence is segmented into morphemes and position information is added to morphemes. Then submorpheme units are obtained by splitting the morpheme units through the maximization of posterior probability terms. The posterior probability terms are computed from the morpheme frequency distribution, the morpheme length distribution, and the morpheme frequency-of-frequency distribution. Finally, the recognition units are obtained by sequentially merging the submorpheme pair with the highest frequency. Computer experiments are conducted using a Korean LVCSR with a 100k word vocabulary and a trigram language model obtained by a 300 million eojeol (word phrase) corpus. The proposed method is shown to reduce the out-of-vocabulary rate to 1.8% and reduce the syllable error rate relatively by 14.0%.

• Journal of Platform Technology
• /
• v.9 no.3
• /
• pp.52-62
• /
• 2021

Quality control is critical at manufacturing sites and is key to predicting the risk of quality defect before manufacturing. However, the reliability of manual quality control methods is affected by human and physical limitations because manufacturing processes vary across industries. These limitations become particularly obvious in domain areas with numerous manufacturing processes, such as the manufacture of major nuclear equipment. This study proposed a novel method for predicting the risk of quality defects by using natural language processing and machine learning. In this study, production data collected over 6 years at a factory that manufactures main equipment that is installed in nuclear power plants were used. In the preprocessing stage of text data, a mapping method was applied to the word dictionary so that domain knowledge could be appropriately reflected, and a hybrid algorithm, which combined n-gram, Term Frequency-Inverse Document Frequency, and Singular Value Decomposition, was constructed for sentence vectorization. Next, in the experiment to classify the risky processes resulting in poor quality, k-fold cross-validation was applied to categorize cases from Unigram to cumulative Trigram. Furthermore, for achieving objective experimental results, Naive Bayes and Support Vector Machine were used as classification algorithms and the maximum accuracy and F1-score of 0.7685 and 0.8641, respectively, were achieved. Thus, the proposed method is effective. The performance of the proposed method were compared and with votes of field engineers, and the results revealed that the proposed method outperformed field engineers. Thus, the method can be implemented for quality control at manufacturing sites.

• Conservation Science in Museum
• /
• v.25
• /
• pp.1-8
• /
• 2021

This study investigated White Porcelain Water Dropper with Openwork Lotus Scroll Design and Eight Trigram Design in Cobalt-blue Underglaze (hereinafter, the "water dropper") in the collection of the National Museum of Korea using computed tomography (CT). A replica was produced to examine both the structure and its original production method. The CT scanning identified no joint lines or pores in the clay, which suggests that the body (the lower part of the water dropper) was shaped in a single piece using a mold and was then matched with a mold-formed lid (the upper part of the water dropper). The inner container of the body portion was roughly trimmed with a bamboo knife so that its upper surface could be securely attached to the bottom of the lid and prevent any leakage in the joined surface. It appears that the inner container for storing water was made first in a cylindrical shape that met the unit of quantity used at the time and could be easily formed by molding. It was transformed into a trapezoid shape during the process of combining it with the lid. A cylindrical inner container was reproduced using silicon 3D printing to compare its capacity with that of the original inner container. The comparison revealed that the reproduced container had a capacity of 152.5㎖, whereas the original container holds approximately 168.6㎖, a figure similar to three hop (around 174㎖) in Joseon-period units of quantity. Since the capacity of the cylindrical inner container corresponds to a known measure from the late Joseon dynasty, it is likely that the water dropper was originally produced to contain a cylindrical inner container.

• Journal of Korean Medical classics
• /
• v.21 no.3
• /
• pp.69-81
• /
• 2008

Gwae(卦氣) theory was one of the main theoretical foundation of yiology in the Han(漢)-dynasty. It was based on the concept that the trigram or hexagram of the book of change corresponds to the seasonal point such as 24 solar terms in one year, so there was so much influence from astronomy and divination system of that time in the development of theoretical principle of Gwae(卦氣) theory. Since Han(漢) Dynasty, the theoretical method such as Gwae(卦氣) theory that correlates the astronomy and divination system with the book of change, had become one of the main academic thoughts throughout the entire history in China, and it was also like that in medicine. Nevertheless there still exists the skeptical sights that Gwae(卦氣) theory was not a part of orthodox yiology, that had been developed by Confucian scholar and had also been recognized as the right path to the study of the book of change. Nowadays because of the new opportunity such as the excavation of the ancient silk script, this kind of controversy has moved on its another step. With regard to this problem, this article will treat the current thoughts about the Gwae(卦氣) theory and think about the substantial basis of each point of view.

• MALSORI
• /
• no.59
• /
• pp.53-68
• /
• 2006

We report the evaluation results of the Korean speech recognition platform called ECHOS. The platform has an object-oriented and reusable architecture so that researchers can easily evaluate their own algorithms. The platform has all intrinsic modules to build a large vocabulary speech recognizer: Noise reduction, end-point detection, feature extraction, hidden Markov model (HMM)-based acoustic modeling, cross-word modeling, n-gram language modeling, n-best search, word graph generation, and Korean-specific language processing. The platform supports both lexical search trees and finite-state networks. It performs word-dependent n-best search with bigram in the forward search stage, and rescores the lattice with trigram in the backward stage. In an 8000-word continuous speech recognition task, the platform with a lexical tree increases 40% of word errors but decreases 50% of recognition time compared to the HTK platform with flat lexicon. ECHOS reduces 40% of recognition errors through incorporation of cross-word modeling. With the number of Gaussian mixtures increasing to 16, it yields word accuracy comparable to the previous lexical tree-based platform, Julius.

• Journal of TMJ Balancing Medicine
• /
• v.7 no.1
• /
• pp.11-16
• /
• 2017

In order to understand Changeologically on the meaning of FCST's TMJ (temporo-mandibular joint) treatment procedure, Yi, Shike, Bi, Gen trigrams concerning the jaw and change by treatment were analyzed from the viewpoint of semiotic context of hexagon and holistic interpretation on disease. Yi is meant by jaw but actually indicates mouth made by maxilla and mandible, and it's characters are related with nourishing by aliment and words. But when we eat and speak in the daily life, jaw does not nourish properly it's own body by bad habit or postures. For the treatment of this ill state, there needs punishment and correction symbolized with Shike. Shike has fourth nine meaning obstacles between the two strong lines in the upper and lower end, and so it has the function of mastication and get rid of the fourth nine metaphorically indicating subluxation of axis using CBA and auxiliary measures of four movement or laughing methods. Bi expresses the achievement and effects of consecutive mastication process implicating normalized manifestation of jaw and its linked spinal function. Gen symbolizes removing selfish motive or partiality in advance and reaches the best state of the saint righteously self-nourishable human being.

• Annual Conference on Human and Language Technology
• /
• 2007.10a
• /
• pp.134-139
• /
• 2007

품사 태깅 시스템에서 규칙 정보와 통계 정보는 상호보완적으로 사용되어 품사 태깅의 성능을 향상시킨다. 하지만, 두 가지 정보로는 품사 태깅의 성능을 향상시키기에는 한계가 있다. 이에 본 논문에서는 어절별 중의성 해소 정보를 이용하여 품사 태깅 시스템의 정확률을 향상시키는 방법에 대해서 기술한다. 통계 정보는 21세기 세종계획의 천만 어절 균형 말뭉치와 태그 부착 말뭉치에서 추출한 trigram 형태의 중의성 어절 및 품사 태그열 출현 빈도 정보를 이용하여 구축하였고, 규칙 정보는 보조용언, 숙어, 관용적 표현 등을 이용하여 구축하였다. 어절별 중의성 해소 정보는 세종 천만 어절 균형 말뭉치의 중의성 어절에서 고빈도 상위 50%에 해당하는 어절을 대상으로 해당 어절의 의미정보와 문맥정보를 고려하여 구축되었고, 이것은 통계 정보를 이용한 품사 태깅 전에 적용되어 분석 후보를 줄여준다. 또한, 학습을 통하여 어절별 중의성 해소 정보를 수정 및 보강하여 잘못된 품사 태깅 결과를 보정해준다. 이와 같이 통계 정보와 규칙 정보를 이용한 품사 태깅 시스템에 고빈도 중의성 어절에 대한 어절별 중의성 해소 정보를 이용함으로써 품사 태깅의 성능을 향상시킬 수 있었다.