• Journal of Intelligence and Information Systems
• /
• v.27 no.3
• /
• pp.175-197
• /
• 2021

Recently, with the development of deep learning technology, research on unstructured data analysis is being actively conducted, and it is showing remarkable results in various fields such as classification, summary, and generation. Among various text analysis fields, text classification is the most widely used technology in academia and industry. Text classification includes binary class classification with one label among two classes, multi-class classification with one label among several classes, and multi-label classification with multiple labels among several classes. In particular, multi-label classification requires a different training method from binary class classification and multi-class classification because of the characteristic of having multiple labels. In addition, since the number of labels to be predicted increases as the number of labels and classes increases, there is a limitation in that performance improvement is difficult due to an increase in prediction difficulty. To overcome these limitations, (i) compressing the initially given high-dimensional label space into a low-dimensional latent label space, (ii) after performing training to predict the compressed label, (iii) restoring the predicted label to the high-dimensional original label space, research on label embedding is being actively conducted. Typical label embedding techniques include Principal Label Space Transformation (PLST), Multi-Label Classification via Boolean Matrix Decomposition (MLC-BMaD), and Bayesian Multi-Label Compressed Sensing (BML-CS). However, since these techniques consider only the linear relationship between labels or compress the labels by random transformation, it is difficult to understand the non-linear relationship between labels, so there is a limitation in that it is not possible to create a latent label space sufficiently containing the information of the original label. Recently, there have been increasing attempts to improve performance by applying deep learning technology to label embedding. Label embedding using an autoencoder, a deep learning model that is effective for data compression and restoration, is representative. However, the traditional autoencoder-based label embedding has a limitation in that a large amount of information loss occurs when compressing a high-dimensional label space having a myriad of classes into a low-dimensional latent label space. This can be found in the gradient loss problem that occurs in the backpropagation process of learning. To solve this problem, skip connection was devised, and by adding the input of the layer to the output to prevent gradient loss during backpropagation, efficient learning is possible even when the layer is deep. Skip connection is mainly used for image feature extraction in convolutional neural networks, but studies using skip connection in autoencoder or label embedding process are still lacking. Therefore, in this study, we propose an autoencoder-based label embedding methodology in which skip connections are added to each of the encoder and decoder to form a low-dimensional latent label space that reflects the information of the high-dimensional label space well. In addition, the proposed methodology was applied to actual paper keywords to derive the high-dimensional keyword label space and the low-dimensional latent label space. Using this, we conducted an experiment to predict the compressed keyword vector existing in the latent label space from the paper abstract and to evaluate the multi-label classification by restoring the predicted keyword vector back to the original label space. As a result, the accuracy, precision, recall, and F1 score used as performance indicators showed far superior performance in multi-label classification based on the proposed methodology compared to traditional multi-label classification methods. This can be seen that the low-dimensional latent label space derived through the proposed methodology well reflected the information of the high-dimensional label space, which ultimately led to the improvement of the performance of the multi-label classification itself. In addition, the utility of the proposed methodology was identified by comparing the performance of the proposed methodology according to the domain characteristics and the number of dimensions of the latent label space.

• Journal of the Daesoon Academy of Sciences
• /
• v.24_2
• /
• pp.231-270
• /
• 2015

This thesis examined Chu Hsi's self-cultivation theory as the representative theory of cultivation in Confucianism, while juxtaposing it to self-cultivation theory in Daesoon thought, concentrating on its similarity and difference. Neo-Confucianism is a scholarship which has wielded a tremendous influence upon East Asia, while functioning as an official scholarship for long period up to Yuan, Ming, and Ching period, after achieving its synthesis by Chu Hsi. After 13th century, Neo-Confucianism has been a representative academic system in Confucianism, and self-cultivation theory was in its center. It suggested various virtues that classical scholars have to equip to fulfill the sainthood and the politics of royal road. The self-cultivation theory of Chu Hsi was developed upon the basis of the theory 'Li-Qi' and the theory of mind. Here, the practice of Geokyung(morally awakened state with a reverent spirit in every moment) and Gyeokmul-tsiji(reaching into the utmost knowledge through investigation of things) formed a nucleus of Chu Hsi's self-cultivation theory. While Geokyung was to reveal one's true nature through cultivation of mind, Gyeokmul-tsiji was to demonstrate the complete use of mind's essence and function. Chu Hsi's self-cultivation theory attempted to realize the unity of heaven and man, and through cultivating self and governing people, it also tried to achieve its ideal of the society of Great Union(Daedong). Daesoon Thought is originated from Sangjenim who has descended to this world as a human being called Jeungsan. He went on his circuit to rectify the disorder of heaven and earth for 9 years to rectify the Three worlds of heaven, earth and human being which were faced with total annihilation due to its rule of mutual conflict, while creating an earthly paradise. Respecting the will of Sangjenim, Dojunim established the foundation of Do through launching 'Meukeukdo' and setting tenet, creed and object so that the cultivators (Doins) could believe and respect the truth of Sangje's great itineration (Daesoon). Among those, the two components of creed, which are four fundamental principles and the three fundamental attitudes are of great account as precept and cardinal point. Through this means, the cultivators of Daesoonjinrihoe aspire to guard against self-deception through transforming the human spirit, to realize earthly immortality through renewing human beings, and to create an earthly paradise through transforming the world. This thesis attempted to compare and analyze the theory of self-cultivation in Neo-Confucianism and Daesoon Thought in the aspect of ground, method, and object of cultivation. First, as for the ground of cultivation, the doctrines of Chu-tzu and Daesoon thought place the essence of cultivation on 'heaven'. Yet, whereas the former postulates Taekeuk (the Great Ultimate) as a principle as well as the heaven of a natural order, Daesoon thought postulates Sangenim as the heaven of superintendence as well as the heaven of a natural order, signified as its equation of Daesoon with circle, along with the unity of Meukeuk (Endlessness) and Taegeuk (the Great Ultimate). Further, the doctrine of Chu-zhu and Daesoon thought is identical in the point that both thoughts see mind as the subject of cultivation, while trying to restore a pure essence. Nevertheless, whereas Neo-Confucianism intends to give scope to ability of the complete use of mind's essence and function, Daesoon Thought sees mind as the essence which is used by spiritual beings and as an organ that heaven, earth and human being rely upon as the center of the universe. In the aspect of method of cultivation, the doctrine of Chuzhu lays emphasis on the rational factor in that it brightens its 'myung-deoki'(bright inner virtue),' while trying to correspond to the law of heaven on the basis of 'Geokyung' and 'Gyeokmul-tsiji.' On the contrary, Daesoon thought lays much emphasis on faith factor in that it aspires for human perfection based on the restoration of conscience by cultivating Daesoonjinri with sincerity, reverence and faith along with 'quieting the heart-mind', 'quieting the body', 'respecting the God of the Ninth Heaven', and 'observing ritual practice on the basis of the faith in Sangjenim. Yet, both thoughts have similarities in that cultivation of body forms the basis and that they attempt to realize their ideals through cultivation in daily life while taking 'Guarding against self-deception' as the key method of self-cultivation.' However, the principle of Chu-zhu can be said to be a voluntary and autonomous practice based on scripture of the saint as well as self-reflection. On the other hand, Daesoon thought reveals certain difference in that it combines faith factor with one's self-effort by concentrating on cultivation under the presence of Sangjenim as the object of belief and the spirits of heaven and earth. In the aspect of object of cultivation, both thoughts share similarities in that the saint and the perfected gentleman with a moral virtue as an ideal image of men in both thoughts attempt to realize each of their 'myung-deok' in human nature as a heavenly mandate while respecting morality. Further, they also share similarity in the point that the desirable characters in both thoughts want to participate in harmonious creation and nurturance. Yet, the perfected gentleman with a moral virtue is also characterized by its aim for a new heaven and earth where there is no mutual conflict but mutual beneficence, by promoting the moral influence and virtue of Sangjenim over one's own virtue, while practising the mutual beneficence of all life through harmonious union of divine beings and human beings.

• Journal of the Korean Institute of Traditional Landscape Architecture
• /
• v.32 no.3
• /
• pp.32-45
• /
• 2014

The study attempts to interpret original location and landscape composition of Yowol-pavilion Garden under the premise that tablet and poetry are important criteria for inference of unique location and landscape composition of a pavilion garden. The study raises the meaning, status, and value of Yowol Pavilion Garden as a cultural asset. The results of the study are as follows. First, Yowol-pavilion Garden was a place where famous Confucius scholars in Joseon Dynasty in 16th Century, including Kim, Kyung-Woo, the owner of the garden, used to share the taste for the arts and poetries with their colleagues. Along with a main characteristic of Yowol Pavilion Garden as a hideout for the Confucius scholars who stayed away from a political turmoil, the new place characteristic of the garden, a bridgehead for the formation of regional identity, was discovered in the record of "Joseon-Hwanyeo-Seungram Honam-Eupji JangSeong-Eupji", As described in "The first creative poetry of Yowol-pavilion", the intention for the creation of Yowol-pavilion Garden and the motive for its landscape composition is interpreted as a space of rivalry where the world, reality and ideals are mixed up. Second, related to outstanding scenic factors and natural phenomena when taking a view from the pavilion, the name of the house 'Yowol', which means 'Greeting the moon rising on the Mt. Wolbong' is the provision of nature and taste for the arts, and is directly connected to the image of leaving the worldly. In other words, the name was identified to be the one that reflected the intention for landscape composition to follow the provision of nature separating from joy and sorrow of the mundane world. Third, as for the location, it was confirmed through "YeongGwang-Soksu-Yeoji-Seungram" that Yowol-pavilion Garden was a place where the person who made the pavilion prepared for relaxation after stepping down from a government post, and literature and various poetry show that it was also a place of outstanding scenic where Yellow-dragon River meandered facing Mt. Wolbong. Especially, according to an interview with a keeper, the visual perception frequency of the nightscape of Yowol-pavilion Garden is the highest when viewing by considering the east, the direction of Yellow-dragon River, as Suksigak[normal angle's view], towards Yowel-pavilion from the keeper's house. In addition, he said that the most beautiful landscape with high perception strength is when the moon came up from the left side of Yowol-pavilion, cuts across the Lagerstroemia india heal in front of Yowol-pavilion, and crosses the meridian between Mt. Wolbong peaks facing Yowol-pavilion. Currently, the exposure of Yowol-pavilion Garden is $SE\;141.2^{\circ}$, which is almost facing southeast. It is assumed that the exposure of Yowol-pavilion Garden was determined considering the optimized direction for appreciating the trace of the moon and the intention of securing the visibility as well as topographic conditions. Furthermore, it is presumed that the exposure of Yowol-pavilion Garden was determined so that the moon is reflected on the water of Yellow-dragon River and the moon and its reflection form a symmetry. Fourth, currently, Yowol-pavilion Garden is divided into 'inner garden sphere' composed of Yowol-pavilion, meeting place of the clan and administration building, and 'outer garden sphere' which is inclusive of entrance space, Crape Myrtle Community Garden and Pine Tree Forest in the back. Further, Yowol-pavilion Garden has been deteriorated as the edge was expanded to 'Small lake[Yong-so] and Gardens of aquatic plants sphere' and recently-created 'Yellow-dragon Pavilion and park sphere'. Fifth, at the time it was first made, Yowol-pavilion Garden was borderless gardens consisting of mountains and water taking a method of occupying a specific space of nearby nature centering around pavilion by embracing landscape viewed from the pavilion, but interpreted current complex landscapes are identified to be entirely different from landscapes of the original due to 'Different Changes', 'Fragmentation' and 'Apart piece' in many parts. Lastly, considering that Yowol-pavilion Garden belongs to the Cultural Properties Protection Zone, though not the restoration to the landscapes of the original described in tablet and literature record, at least taking a measure from the aspect of land use for minimizing adverse effect on landscape and visual damage is required.