• Title/Summary/Keyword: 베이스

Search Result 4,380, Processing Time 0.03 seconds

Probing Sentence Embeddings in L2 Learners' LSTM Neural Language Models Using Adaptation Learning

  • Kim, Euhee
    • Journal of the Korea Society of Computer and Information
    • /
    • v.27 no.3
    • /
    • pp.13-23
    • /
    • 2022
  • In this study we leveraged a probing method to evaluate how a pre-trained L2 LSTM language model represents sentences with relative and coordinate clauses. The probing experiment employed adapted models based on the pre-trained L2 language models to trace the syntactic properties of sentence embedding vector representations. The dataset for probing was automatically generated using several templates related to different sentence structures. To classify the syntactic properties of sentences for each probing task, we measured the adaptation effects of the language models using syntactic priming. We performed linear mixed-effects model analyses to analyze the relation between adaptation effects in a complex statistical manner and reveal how the L2 language models represent syntactic features for English sentences. When the L2 language models were compared with the baseline L1 Gulordava language models, the analogous results were found for each probing task. In addition, it was confirmed that the L2 language models contain syntactic features of relative and coordinate clauses hierarchically in the sentence embedding representations.

Analysis of Musical Characteristics and Changes in Different Periods on Yoon-Sang's Music (윤상의 곡에 나타난 음악적 특징과 시대별 변화)

  • Park, Ji-Eun;Chung, Jae-Youn
    • Journal of Korea Entertainment Industry Association
    • /
    • v.15 no.1
    • /
    • pp.63-73
    • /
    • 2021
  • This study aims to analyze music of Yoon-sang, as a part of musical research, which is the most fundamental approach among academic studies on Korean popular music. Yoon-Sang is a representative composer, who has gone through the 1980s to the present. The result of analysis of 21 songs created by Yoon-Sang showed that his songs are mostly characterized by tonal music, in which chord relationships develop focusing on keynotes. The reason why his music does not sound uniform pursuing stability is he properly added the progression of chromatic chords, based on diatonic chords and melodies. Dominant 7th chord and diminished 7th chord are used the most among diverse techniques adding chromatic colors. Along with these chords, chromatic intervals are used not only in chord progression but also in melodies. The successive, ascending or descending movement of the base line is his common composition and arrangement technique revealed in every song. One of formal changes with the stream of the times is that the number of measured in the pre-chorus and interlude that were of great importance in his songs of the 1990s decreased over time. With regard to harmonic changes, whereas modulation between parts was applied to his 2 songs created in the 2010s. Yoon-Sang's music had one strong tonality overall, but his music began to have more than two tonalities starting the 2010s, and this is a big variation in his music.

Study of Spectral Doppler Waveform Interpretation and Nomenclature in Peripheral Artery (말초 동맥 분광 도플러 파형 해석 및 명명법에 대한 고찰)

  • Ji, Myeong-Hoon;Seoung, Youl-Hun
    • Journal of the Korean Society of Radiology
    • /
    • v.16 no.5
    • /
    • pp.649-660
    • /
    • 2022
  • In 1959, Satomura used spectral Doppler ultrasound to express the velocity of red blood cells according to time change, and Kato defined a zero-base line that could tell the direction of blood flow, making it possible to know the direction of blood flow. This became the basis for the widely used classifications of Triphasic, Biphasic, and Monophasic. However, the above classification has limitations that confuse users with the meaning and timing of use in a clinical environment. As a result, the American Society for Vascular Medicine (SVM) and the Society for Vascular Ultrasound (SVU) A consensus document on Doppler waveform analysis was declared by the joint committee. This study tried to review this consensus and to suggest nomenclature and modifiers that can be used in the domestic vascular ultrasound clinical field. The joint committee formed by SVM and SVU recommended that the use of the triphasic waveform and the biphasic waveform be used as a multiphasic waveform rather than being used due to the ambiguity of interpretation. In addition, it was agreed to name the hybrid-type waveform, which is a monophasic and high-resistance waveform, which has always been a problem of interpretation in a clinical environment, as an intermediate resistive waveform. In addition, in order to increase the communication efficiency between the interpreter and the sonographer, waveform analysis was classified into a main descriptor and a modifier, and it was recommended to use a single nomenclature by unifying various synonyms. It is expected that this literature review will provide accurate arterial spectral Doppler waveform interpretation and an agreed-upon nomenclature to radiologists performing vascular ultrasound examination in clinical practice, and will be utilized as basic data that can contribute to the improvement of public health.

Makeup transfer by applying a loss function based on facial segmentation combining edge with color information (에지와 컬러 정보를 결합한 안면 분할 기반의 손실 함수를 적용한 메이크업 변환)

  • Lim, So-hyun;Chun, Jun-chul
    • Journal of Internet Computing and Services
    • /
    • v.23 no.4
    • /
    • pp.35-43
    • /
    • 2022
  • Makeup is the most common way to improve a person's appearance. However, since makeup styles are very diverse, there are many time and cost problems for an individual to apply makeup directly to himself/herself.. Accordingly, the need for makeup automation is increasing. Makeup transfer is being studied for makeup automation. Makeup transfer is a field of applying makeup style to a face image without makeup. Makeup transfer can be divided into a traditional image processing-based method and a deep learning-based method. In particular, in deep learning-based methods, many studies based on Generative Adversarial Networks have been performed. However, both methods have disadvantages in that the resulting image is unnatural, the result of makeup conversion is not clear, and it is smeared or heavily influenced by the makeup style face image. In order to express the clear boundary of makeup and to alleviate the influence of makeup style facial images, this study divides the makeup area and calculates the loss function using HoG (Histogram of Gradient). HoG is a method of extracting image features through the size and directionality of edges present in the image. Through this, we propose a makeup transfer network that performs robust learning on edges.By comparing the image generated through the proposed model with the image generated through BeautyGAN used as the base model, it was confirmed that the performance of the model proposed in this study was superior, and the method of using facial information that can be additionally presented as a future study.

Integrated receptive field diversification method for improving speaker verification performance for variable-length utterances (가변 길이 입력 발성에서의 화자 인증 성능 향상을 위한 통합된 수용 영역 다양화 기법)

  • Shin, Hyun-seo;Kim, Ju-ho;Heo, Jungwoo;Shim, Hye-jin;Yu, Ha-Jin
    • The Journal of the Acoustical Society of Korea
    • /
    • v.41 no.3
    • /
    • pp.319-325
    • /
    • 2022
  • The variation of utterance lengths is a representative factor that can degrade the performance of speaker verification systems. To handle this issue, previous studies had attempted to extract speaker features from various branches or to use convolution layers with different receptive fields. Combining the advantages of the previous two approaches for variable-length input, this paper proposes integrated receptive field diversification that extracts speaker features through more diverse receptive field. The proposed method processes the input features by convolutional layers with different receptive fields at multiple time-axis branches, and extracts speaker embedding by dynamically aggregating the processed features according to the lengths of input utterances. The deep neural networks in this study were trained on the VoxCeleb2 dataset and tested on the VoxCeleb1 evaluation dataset that divided into 1 s, 2 s, 5 s, and full-length. Experimental results demonstrated that the proposed method reduces the equal error rate by 19.7 % compared to the baseline.

A Study on Korea Land Use Information System Zoning Data Maintenance Plan (국토이용정보체계 용도지역지구 데이터 정비방안)

  • Lee, Se-won
    • Journal of Cadastre & Land InformatiX
    • /
    • v.51 no.2
    • /
    • pp.51-72
    • /
    • 2021
  • The purpose of this study is to explain the types and causes of errors in zoning data that occur in the topographic map notification procedure, and to prepare a data maintenance plan. In Korea, like the United States, law-based land use regulation is dominant. In other words, according to the land use regulation method in the Act, the government designates zoning for all lots in the country, and landowners check the land use regulations of their land through the Korea Land use Information System. The land use plan confirmation document is important land information that affects the results of administrative dispositions such as land transactions between individuals or permission for development activities. However, there are data errors that occur during the current topographic map notification procedure and data construction process. Therefore, four local governments that can verify data by type were selected in consideration of local government conditions. A number of errors are first, errors in data construction and management in the Korea Land use Information System, and second, errors in lack of expertise that occur while the local government officials maintain data. Third, it was analyzed as an error from the relationship between the serial cadastral map and the zoning DB. Based on the above results, it is hoped that the results of this study will be reflected in the establishment of the KLIP and the reform of the legal system, which is currently underway after the establishment of the 「3rd the Korea Land use Information System Construction Plan」.

Improved Estimation of Hourly Surface Ozone Concentrations using Stacking Ensemble-based Spatial Interpolation (스태킹 앙상블 모델을 이용한 시간별 지상 오존 공간내삽 정확도 향상)

  • KIM, Ye-Jin;KANG, Eun-Jin;CHO, Dong-Jin;LEE, Si-Woo;IM, Jung-Ho
    • Journal of the Korean Association of Geographic Information Studies
    • /
    • v.25 no.3
    • /
    • pp.74-99
    • /
    • 2022
  • Surface ozone is produced by photochemical reactions of nitrogen oxides(NOx) and volatile organic compounds(VOCs) emitted from vehicles and industrial sites, adversely affecting vegetation and the human body. In South Korea, ozone is monitored in real-time at stations(i.e., point measurements), but it is difficult to monitor and analyze its continuous spatial distribution. In this study, surface ozone concentrations were interpolated to have a spatial resolution of 1.5km every hour using the stacking ensemble technique, followed by a 5-fold cross-validation. Base models for the stacking ensemble were cokriging, multi-linear regression(MLR), random forest(RF), and support vector regression(SVR), while MLR was used as the meta model, having all base model results as additional input variables. The results showed that the stacking ensemble model yielded the better performance than the individual base models, resulting in an averaged R of 0.76 and RMSE of 0.0065ppm during the study period of 2020. The surface ozone concentration distribution generated by the stacking ensemble model had a wider range with a spatial pattern similar with terrain and urbanization variables, compared to those by the base models. Not only should the proposed model be capable of producing the hourly spatial distribution of ozone, but it should also be highly applicable for calculating the daily maximum 8-hour ozone concentrations.

Effect of Babassu Oil on the Improvement of Damaged Hair (바바수오일이 손상된 모발 개선에 미치는 영향)

  • Kim, Ju-Sub;Uhm, Sang Jun
    • Journal of the Korean Applied Science and Technology
    • /
    • v.39 no.3
    • /
    • pp.471-478
    • /
    • 2022
  • This study attempted to find out the improvement effect the improvement effect of Babassu oil on damaged hair by damaged hair by applying a hair-improving formulation with Babassu oil to damaged hair after manufacturing. As the experimental raw material, the contents of babassu oil were changed to 0 g, 3 g 6 g, and 9 g and added to the perm base agent. The prepared formulation was applied to 8 levels of bleached sample hair. Each sample and damaged hair were measured and compared and analyzed. As for the measurement method, tensile strength, absorbance using methylene blue, and gloss were measured to know the effect of improving damage hair Statistical analysis was conducted for the reliability of the research results. As a result of the research, it was found that the tensile strength was higher than that of the damaged hair in all samples applied by adding babassu oil. As a result of absorbance analysis using methylene blue, it was found that absorbance was decreased in all samples compared to damaged hair. As a result of the gloss measurement, it was found that the gloss was increased in all samples compared to the damaged hair. In conclusion, it was found that babassu oil had an improvement effect on damaged hair.

A Study on the Digital Craft Convergence Cases Using 3D Printing Technology (3D Printing 기술을 활용한 Digital Craft 융합적 사례 연구)

  • Jang, Jisu;Chung, Jeanhun
    • The Journal of the Institute of Internet, Broadcasting and Communication
    • /
    • v.22 no.3
    • /
    • pp.105-110
    • /
    • 2022
  • In line with the trend of the 4th Industrial Revolution, we tried to study the scalability of the craft sector in the art field in the wake of the convergence of new fields through the combination of different fields. To this end, a craft case using 3D printing technology was analyzed based on the theoretical background of the definition and value of digital craft. As a result of the study, craft works using 3D printing technology could be divided into four forms according to the fusion expression method. First, the mixed type was most characterized by harmonious production without significantly deviating from the traditional craft type. Second, the component type was produced in the form in which the characteristics of digital technology, which has advantages of certain shapes and detailed work, were best exhibited. Third, the structural type became a work with analog sensibility by adding traditional craft techniques based on the results output from a 3D printer. Finally, the connection type was the work that showed the greatest glimpse of originality and uniqueness among the analysis cases of this study. As digital technology is positively widely used, future studies will also deal with effective work directions.

TeGCN:Transformer-embedded Graph Neural Network for Thin-filer default prediction (TeGCN:씬파일러 신용평가를 위한 트랜스포머 임베딩 기반 그래프 신경망 구조 개발)

  • Seongsu Kim;Junho Bae;Juhyeon Lee;Heejoo Jung;Hee-Woong Kim
    • Journal of Intelligence and Information Systems
    • /
    • v.29 no.3
    • /
    • pp.419-437
    • /
    • 2023
  • As the number of thin filers in Korea surpasses 12 million, there is a growing interest in enhancing the accuracy of assessing their credit default risk to generate additional revenue. Specifically, researchers are actively pursuing the development of default prediction models using machine learning and deep learning algorithms, in contrast to traditional statistical default prediction methods, which struggle to capture nonlinearity. Among these efforts, Graph Neural Network (GNN) architecture is noteworthy for predicting default in situations with limited data on thin filers. This is due to their ability to incorporate network information between borrowers alongside conventional credit-related data. However, prior research employing graph neural networks has faced limitations in effectively handling diverse categorical variables present in credit information. In this study, we introduce the Transformer embedded Graph Convolutional Network (TeGCN), which aims to address these limitations and enable effective default prediction for thin filers. TeGCN combines the TabTransformer, capable of extracting contextual information from categorical variables, with the Graph Convolutional Network, which captures network information between borrowers. Our TeGCN model surpasses the baseline model's performance across both the general borrower dataset and the thin filer dataset. Specially, our model performs outstanding results in thin filer default prediction. This study achieves high default prediction accuracy by a model structure tailored to characteristics of credit information containing numerous categorical variables, especially in the context of thin filers with limited data. Our study can contribute to resolving the financial exclusion issues faced by thin filers and facilitate additional revenue within the financial industry.