• The KIPS Transactions:PartB
• /
• 제8B권6호
• /
• pp.685-692
• /
• 2001

This paper describes how to disambiguate the aspectual meaning of Japanese expression "-te iru" in Japanese-Korean machine translation Due to grammatical similarities of both languages, almost all Japanese- Korean MT systems have been developed under the direct MT strategy, in which the lexical disambiguation is essential to high-quality translation. Japanese has a progressive aspectual marker “-te iru" which is difficult to translate into Korean equivalents because in Korean there are two different progressive aspectual markers: "-ko issta" for "action progressive" and "-e issta" for "state progressive". Moreover, the aspectual system of both languages does not quite coincide with each other, so the Korean progressive aspect could not be determined by Japanese meaning of " te iru" alone. The progressive aspectural meaning may be parially determined by the meaning of predicates and also the semantic meaning of predicates may be partially reshicted by adverbials, so all Japanese predicates are classified into five classes : the 1nd verb is used only for "action progrssive",2nd verb generally for "action progressive" but occasionally for "state progressive", the 3rd verb only for "state progressive", the 4th verb generally for "state progressive", but occasIonally for "action progressive", and the 5th verb for the others. Some heuristic rules are defined for disambiguation of the 2nd and 4th verbs on the basis of adverbs and abverbial phrases. In an experimental evaluation using more than 15,000 sentances from "Asahi newspapers", the proposed method improved the translation quality by about 5%, which proves that it is effective in disambiguating "-te iru" for Japanese-Korean machine translation.translation quality by about 5%, which proves that it is effective in disambiguating "-te iru" for Japanese-Korean machine translation.anslation.

• Journal of the Korean earth science society
• /
• 제43권6호
• /
• pp.691-711
• /
• 2022

The state-of-the-art Earth system model as a virtual Earth is required for studies of current and future climate change or climate crises. This complex numerical model can account for almost all human activities and natural phenomena affecting the atmosphere of Earth. The Unified Model (UM) from the United Kingdom Meteorological Office (UK Met Office) is among the best Earth system models as a scientific tool for studying the atmosphere. However, owing to the expansive numerical integration cost and substantial output size required to maintain the UM, individual research groups have had to rely only on supercomputers. The limitations of computer resources, especially the computer environment being blocked from outside network connections, reduce the efficiency and effectiveness of conducting research using the model, as well as improving the component codes. Therefore, this study has presented detailed guidance for installing a new version of the UM on high-performance parallel computers (Linux clusters) owned by individual researchers, which would help researchers to easily work with the UM. The numerical integration performance of the UM on Linux clusters was also evaluated for two different model resolutions, namely N96L85 (1.875° ×1.25° with 85 vertical levels up to 85 km) and N48L70 (3.75° ×2.5° with 70 vertical levels up to 80 km). The one-month integration times using 256 cores for the AMIP and CMIP simulations of N96L85 resolution were 169 and 205 min, respectively. The one-month integration time for an N48L70 AMIP run using 252 cores was 33 min. Simulated results on 2-m surface temperature and precipitation intensity were compared with ERA5 re-analysis data. The spatial distributions of the simulated results were qualitatively compared to those of ERA5 in terms of spatial distribution, despite the quantitative differences caused by different resolutions and atmosphere-ocean coupling. In conclusion, this study has confirmed that UM can be successfully installed and used in high-performance Linux clusters.

• Journal of Internet Computing and Services
• /
• 제16권1호
• /
• pp.67-74
• /
• 2015

In recent years, frequent itemset mining for considering the importance of each item has been intensively studied as one of important issues in the data mining field. According to strategies utilizing the item importance, itemset mining approaches for discovering itemsets based on the item importance are classified as follows: weighted frequent itemset mining, frequent itemset mining using transactional weights, and utility itemset mining. In this paper, we perform empirical analysis with respect to frequent itemset mining algorithms based on transactional weights. The mining algorithms compute transactional weights by utilizing the weight for each item in large databases. In addition, these algorithms discover weighted frequent itemsets on the basis of the item frequency and weight of each transaction. Consequently, we can see the importance of a certain transaction through the database analysis because the weight for the transaction has higher value if it contains many items with high values. We not only analyze the advantages and disadvantages but also compare the performance of the most famous algorithms in the frequent itemset mining field based on the transactional weights. As a representative of the frequent itemset mining using transactional weights, WIS introduces the concept and strategies of transactional weights. In addition, there are various other state-of-the-art algorithms, WIT-FWIs, WIT-FWIs-MODIFY, and WIT-FWIs-DIFF, for extracting itemsets with the weight information. To efficiently conduct processes for mining weighted frequent itemsets, three algorithms use the special Lattice-like data structure, called WIT-tree. The algorithms do not need to an additional database scanning operation after the construction of WIT-tree is finished since each node of WIT-tree has item information such as item and transaction IDs. In particular, the traditional algorithms conduct a number of database scanning operations to mine weighted itemsets, whereas the algorithms based on WIT-tree solve the overhead problem that can occur in the mining processes by reading databases only one time. Additionally, the algorithms use the technique for generating each new itemset of length N+1 on the basis of two different itemsets of length N. To discover new weighted itemsets, WIT-FWIs performs the itemset combination processes by using the information of transactions that contain all the itemsets. WIT-FWIs-MODIFY has a unique feature decreasing operations for calculating the frequency of the new itemset. WIT-FWIs-DIFF utilizes a technique using the difference of two itemsets. To compare and analyze the performance of the algorithms in various environments, we use real datasets of two types (i.e., dense and sparse) in terms of the runtime and maximum memory usage. Moreover, a scalability test is conducted to evaluate the stability for each algorithm when the size of a database is changed. As a result, WIT-FWIs and WIT-FWIs-MODIFY show the best performance in the dense dataset, and in sparse dataset, WIT-FWI-DIFF has mining efficiency better than the other algorithms. Compared to the algorithms using WIT-tree, WIS based on the Apriori technique has the worst efficiency because it requires a large number of computations more than the others on average.

• Science of Emotion and Sensibility
• /
• 제15권4호
• /
• pp.553-564
• /
• 2012

Tangible user interfaces have been developed in the area of Human-Computer Interaction for the last decades, however, the applied domains recently have been extended into the product design and interactive art. Tangible User Interfaces are the combination of digital information and physical objects or environments, thus they provide tangible and intuitive interaction as input and output devices, often combined with Augmented Reality. The research developed a design guideline for tangible user interfaces based on key properties of tangible user interfaces defined previously in five representative research: Tangible Interaction, Intuitiveness and Convenience, Expressive Representation, Context-aware and Spatial Interaction, and Social Interaction. Using the guideline emphasizing user interaction, this research evaluated installation in a science museum in terms of the applied characteristics of tangible user interfaces. The selected 15 installations which were evaluated are to educate visitors for science by emphasizing manipulation and experience of interfaces in those installations. According to the input devices, they are categorized into four Types. TUI properties in Type 3 installation, which uses body motions for interaction, shows the highest score, where items for context-aware and spatial interaction were highly rated. The context-aware and spatial interaction have been recently emphasized as extended properties of tangible user interfaces. The major type of installation in the science museum is equipped with buttons and joysticks for physical manipulation, thus multimodal interfaces utilizing visual, aural, tactile senses etc need to be developed to provide more innovative interaction. Further, more installation need to be reconfigurable for embodied interaction between users and the interactive space. The proposed design guideline can specify the characteristics of tangible user interfaces, thus this research can be a basis for the development and application of installation involving more TUI properties in future.

• Cartoon and Animation Studies
• /
• 통권22호
• /
• pp.47-63
• /
• 2011

Recently, contemporary cultural contents have been shown its diversity changes followed by the birth of new media platforms with consumers' new needs in the global market. Also, developments of Internet and computer system networks are the main contributors of making this changes happened rapidly. This study aims to know that how to usefully use those new media platforms through the great example of stop-motion animation music video by analyzing its production and marketing process. The music video production had been focused to be completed with high quality by adjusting the production process economically in spite of the relatively short period(less than one month)from its crank-up to the deadline. Because the production was planned that main characters lead the whole story, the creative team had been tried to reduce the production hours by commonly use the same mold when they make original clay models by collecting the similarities of characters' appearances. By using CG technic, could overcome the visual monotonous from the similarities which inferred above. Also, the repeated rhythm in the music video, the similar scenes of backgrounds were commonly used by copy of the original scene. At the point of directing, the creative team considered both economical and art aspects for the quality work. In details, they divided the scenes into foreground and background, and removed unnecessary parts to save the production hours and budget but make depth of fields in the scenes. Except the viral marketing purposes, was searching for the methods to compensate the production cost. For this, the characters in the music video dressed the same T-shirts which are world-cup logo on, and those were designed for the sale after released the music video. Even the result of the sales was not enough to satisfied, it was estimated a great attempt to the domestic animation industry.

• Journal of Intelligence and Information Systems
• /
• 제22권2호
• /
• pp.127-142
• /
• 2016

Deep learning model is a kind of neural networks that allows multiple hidden layers. There are various deep learning architectures such as convolutional neural networks, deep belief networks and recurrent neural networks. Those have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. Among those architectures, convolutional neural networks and recurrent neural networks are classified as the supervised learning model. And in recent years, those supervised learning models have gained more popularity than unsupervised learning models such as deep belief networks, because supervised learning models have shown fashionable applications in such fields mentioned above. Deep learning models can be trained with backpropagation algorithm. Backpropagation is an abbreviation for "backward propagation of errors" and a common method of training artificial neural networks used in conjunction with an optimization method such as gradient descent. The method calculates the gradient of an error function with respect to all the weights in the network. The gradient is fed to the optimization method which in turn uses it to update the weights, in an attempt to minimize the error function. Convolutional neural networks use a special architecture which is particularly well-adapted to classify images. Using this architecture makes convolutional networks fast to train. This, in turn, helps us train deep, muti-layer networks, which are very good at classifying images. These days, deep convolutional networks are used in most neural networks for image recognition. Convolutional neural networks use three basic ideas: local receptive fields, shared weights, and pooling. By local receptive fields, we mean that each neuron in the first(or any) hidden layer will be connected to a small region of the input(or previous layer's) neurons. Shared weights mean that we're going to use the same weights and bias for each of the local receptive field. This means that all the neurons in the hidden layer detect exactly the same feature, just at different locations in the input image. In addition to the convolutional layers just described, convolutional neural networks also contain pooling layers. Pooling layers are usually used immediately after convolutional layers. What the pooling layers do is to simplify the information in the output from the convolutional layer. Recent convolutional network architectures have 10 to 20 hidden layers and billions of connections between units. Training deep learning networks has taken weeks several years ago, but thanks to progress in GPU and algorithm enhancement, training time has reduced to several hours. Neural networks with time-varying behavior are known as recurrent neural networks or RNNs. A recurrent neural network is a class of artificial neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Unlike feedforward neural networks, RNNs can use their internal memory to process arbitrary sequences of inputs. Early RNN models turned out to be very difficult to train, harder even than deep feedforward networks. The reason is the unstable gradient problem such as vanishing gradient and exploding gradient. The gradient can get smaller and smaller as it is propagated back through layers. This makes learning in early layers extremely slow. The problem actually gets worse in RNNs, since gradients aren't just propagated backward through layers, they're propagated backward through time. If the network runs for a long time, that can make the gradient extremely unstable and hard to learn from. It has been possible to incorporate an idea known as long short-term memory units (LSTMs) into RNNs. LSTMs make it much easier to get good results when training RNNs, and many recent papers make use of LSTMs or related ideas.

• Korean Journal of Culture and Social Issue
• /
• 제12권5호_spc
• /
• pp.81-98
• /
• 2006

The objective of this research is to do a cultural comparison on the daily lives of the children of Korea, Japan and China. To achieve this objective, the questionnares were distributed to the 2940 mothers of children from the ages of 3 to 6 in the countries of Korea, Japan and China. The target audience consisted of 941 mothers living in Seoul and Kyunggi area for Korea, 1007 mothers living in Tokyo for Japan, and 992 mothers living in Beijing for China. As a result of the research, we found out that firstly, although children in general got up anytime between 7:00am to 9:00am and went to bed between 8:00pm and 11:00pm, 61.5% of the Korean children went to bed after 10pm and 16.8% after 11pm. Besides that, we found that compared to 3.51% of Korean children who got up before 6am, 13.41% of Japanese children and 17.24% of Chinese children got up before 6:00am. So we could see that the Korean children got up later and went to bed later than their Japanese and Chinese counterpart. This pattern could also be seen in the average rising time and bed time. Korean children went to bed at 10:00pm and woke up at 7:75am whereas the Japanese children went to bed at 9:28pm and woke up at 7:39am, and the Chinese children went to bed at 9:05pm and woke up at 7:05am. The average sleeping hours for Japanese children was 10.12 hours, 9.50 hours for the Chinese and 9.75 hours for the Korean. As a result, we could see that the Korean children went to bed later, got up later and slept fewer hours than their Japanese and Chinese counterparts. Also, since the rising time and bedtime of the Korean children was later than those of the Chinese and Japanese counterparts, the former s' breakfast and dinner time was also much later. Secondly, we looked at the time children went off to and came back from institutes such as kindergarten and child care centers. The Chinese were earliest at going with average attendance at 7:83am, the Japanese came next at 8:59am and the Korean children were last at 8:90am, whereas the Japanese came first in coming back home at 3:36pm, Korean next at 3:91pm and the Chinese last at 5:46pm. Next when we looked at the hours spent at the kindergartens and child care centers, Japan spent 6.76 hours, Korea 7.01 hours and China spent the longest hours with 9.63 hours. Excluding China where all preschool institutes are centralized into kindergartens, we nest looked at time children went to and came back from the institutes as well as the time spent there. In the case of kindergarten, there was not much difference but in the case of child care centers, the Japanese children went to the child care centers mach earlier and came home later than the Korean children. Also, the time spent at the child care center was much longer for the Japanese than the Korean children. This fact coincides with the Korean mothers' number one wish to the kindergartens and child care centers i.e. for the institutes to prolong their school hours. Thus, the time spent at child care centers for Korea was 7.75 hours, 9.39 hours for Japan and 9.63 hours for China. The time for Korea was comparatively much shorter than that of Japan and China but if we consider the fact that 50% of the target audience was working mothers, we could easily presume that the working parents who usually use the child care centers would want the child care centers to prolong the hours looked after their children. Besides this, the next most wanted wish mothers have towards the child care centers and kindergartens was for those institutes to "look after their children when sick". This item showed high marks in all three countries, and the marks in Korea was especially higher when compared to Japan and China. Thirdly, we looked at the private extracurricular activities of the children. We found that 72.6% of the Korean children, 61.7% of the Japanese children, and 64.6% of the Chinese children were doing private extracurricular activities after attending kindergarten or day care centers. Amongst the private extracurricular activities done by Korean children, the most popular one was worksheet with 51.9% of the children doing it. Drawing (15.20%) and English (11.6%) came next. Swimming (21.95%) was the most popular activity for Japan, with English (17.48%), music (15,79%) and sports (14.70%) coming next. For China, art (30.95%) was first with English (22.08%) and music (19.96%) following next. All three countries had English as the most popular activity related to art and physical activities after school hours, but the rate for worksheet studies was much higher for Korea compared to Japan China. The reason Koreans universally use worksheet in because the parents who buy the worksheet are mothers who have easy access to advertisement or salespeople selling those products. The price is also relatively cheap, the worksheet helps the children to grow the basic learning ability in preparation for elementary school, and it is thought to help the children to build the habit of studying everyday. Not only that but it is estimated that the worksheet education is being conducted because parents can share the responsibility of the children's learning with the worksheet-teacher who make home visits. Looking at the expenses spent on private extracurricular activities as compared to income, we found that China spent 5% of income for activities outside of regular education, Korea 3% and Japan 2%. Fourthly, we looked at the amount of time children spent on using multimedia. The majority of the children in Korea, Japan and China watch television almost every day. In terms of video games, the Japanese children played the games the most, with Korea and China following next. The Korean children used the computer the most, with Japan and China next. The Korean children used about 21.17% of their daily time on computers which is much more than the Japanese who used 20.62% of their time 3 or 4 times a week, or the Chinese. The Chinese children were found to use considerably less time on multimedia compared to the Korean of Japanese.

• Journal of Sasang Constitutional Medicine
• /
• 제8권1호
• /
• pp.101-186
• /
• 1996

The clinical application of constitutional Diagnosis is the most important part of Sasang constitutional medicine. It has been studied in various way. However, the study of morphologic characteristics on the face is applied for the first time. For quantitative analysis of the correlation between the sasang constitution and the shape of the face, the head-facial part of 170 cases were measured by Martin's measurement and analysis of a) the measurement value of height and the component ratio from the Gnathion to each part of face by constitution. b) the measurement value of depth and the component ratio from T-projected to each part of the face by constitution. c) the measurement value of breadth and component ratio between each parts of the facial breadth by constitution. d) the ratio of square on every part of face by constitution. e) the characteristics on each part of the face by constitution. f) the contour line of the forehead. g) the result of discriminant analysis about the constitution. Authors obtained the results from the study as follows; 1. The characteristics of Taeum-IN (1) The measurement value of Height, Breadth, T-Projected had a tendency to maximum value in general. (2) The value of lower opthal height and the square of lower opthal part was maximum. (3) The value of Pronasal T-projected length and Subnasal T-projected length was minimum, so Taeum-In has characteristics of depression in middle face, nasal part. (4) In the ratio of Breadth, T-Projected, T-Projected was minimum. (5) It was maximum that the square of nose, Alare, Middle face, Lower face and it was minimum that the square of eye. The square of nose, Alare, Middle facc, Lower face was maximum and the square of eye was minimum. (6) The curvature of the eyebrow was minimum. (7) The projection of jaw (Pogonion T-projection length) was maximum. (8) The breadth of eye was minimum. (9) There was a tendency that the projection of the forehead to the right in general. 2. The characteristics of Soeum-In (1) In all cases of projected length the measurement value was minimum. (2) The value of lower opthal height and the square of lower opthal part was minimum. (3) By the Pupulare T-projected length, the value of Pronasal T-projected length and Subnasal T-projected length was minimum, so the Soeum In's face shape is flat. (4) The square of eye, mouth, forehead was maximum and the square of nose, Alare, Middle face, Lower face was minimum. (5) The curvature of the eyebrow was maximum. (6) The projection of mouth was minimum. (7) The jaw was flat. (8) The breadth of eye was maximum. (9) There was a tendency that the projection of the forehead to the left in general. 3. The characteristics of Soyang-In. (1) In most cases of 고경 length the measurement value was minimum. (2) By the Pupulare T-projected length, each ratio of projected length was maximum, so the Soyang-In's face shape has many protrusions (3) In the ratio of Breadth, T-Projected, T-Projected was maximum. (4) The square of mouth was minimum. (5) The inclination of the forehead was minimum. (6) The projection of mouth was maximum. (7) The breadth of eye was minimum. (8) There was a tendency that the projection of the forehead to the left in general. (9) The middle face was protruded. 4. Discriminant about the constitution. According to the result of discriminant, the accuracy probability of discriminant was 85.58% in total and Taeum-In was 90.5%, Soeum-In was 70.8%, Soyang-In was 89.5%. The accuracy probability of discriminant about 3 constitutional group increased by 49.03% than the accident probility 36.55% 5. Suggestion (1) The study which gather and analysis the data should be continued. (2) The study which subdivide the characteristics of each part of the face by the constitution should be continued. (3) The analysis method about Moire should be supplement. (4) The study about the morphologic characteristics of the whole body should be continued. (5) Computer program of constitution diagnosis should be developed. (6) To increase utility of this method, the measurement should be automation.