• Journal of the Korean Institute of Traditional Landscape Architecture
• /
• v.36 no.3
• /
• pp.26-36
• /
• 2018

This study aims to examine the characteristics of landscape through the analysis of location and the landscape of Gugok while also conducting the empirical study through the literature review, field study, and digital analysis of the Okgung Gugok. Oksan Gugok is a set of songs set in Ogsan Creek(玉山川)or Jagyese Creek(紫溪川, 紫玉山), which flows in front of the Oksan Memorial Hall(李彦迪), which is dedicated to the Lee Eong-jeok (李彦迪). We first ascertained the location and configuration of Oksan Gogok. Second, we confirmed the accurate location of Oksan Gogok by utilizing the digital topographic map of Oksan Gogok which was submitted by Google Earth Pro and Geographic Information Center as well as the length of the longitude of the gravel measured by the Trimble Juno SB GPS. Through the study of the literature and the field investigation, The results of the study are as follows. First, Yi Eonjeok was not a direct composer of Oksan Gugok, nor did he produce "Oksan Gugokha(Music)". Lee Ia-sung(李野淳), the ninth Youngest Son of Tweo-Kye, Hwang Lee, visited the "Oksan Gugokha" in the spring of 1823(Sunjo 23), which was the 270th years after the reign of Yi Eonjeok. At this time, receiving the proposal of Ian Sung, Lee Jung-eom(李鼎儼), Lee Jung-gi(李鼎基), and Lee Jung-byeong(李鼎秉), the descendants of Ian Sung set up a song and created Oksan Gugok Music. And the Essay of Oksan Travel Companions writted by Lee Jung-gi turns out being a crucial data to describe the situation when setting up the Ok-San Gugok. Second, In the majority of cases, Gogok Forest is a forest managed by a Confucian Scholar, not run by ordinary people. The creation of "Oksan Bugok Music" can be regarded as an expression of pride that the descendants of Yi Eonjeok and Lee Hwang, and next generation of several Confucian scholars had inherited traditional Neo-Confucian. Third, Lee Jung-eom's "Oksan Donghaengki" contains a detailed description of the "Oksan Gugokha" process and the process of creating a song. Fourth, We examined the location of one to nine Oksan songs again. In particular, eight songs and nine songs were located at irregular intervals, and eight songs were identified as $36^{\circ}01^{\prime}08.60^{{\prime}{\prime}}N$, $129^{\circ}09^{\prime}31.20^{{\prime}{\prime}}E$. Referring to the ancient kingdom of Taojam, the nine-stringed Sainam was unbiased as a lower rock where the two valleys of the East West congregate. The location was estimated at $36^{\circ}01^{\prime}19.79^{{\prime}{\prime}}N$, $129^{\circ}09^{\prime}30.26^{{\prime}{\prime}}E$. Fifth, The landscape elements and landscapes presented in Lee Jung-eom's "Oksan Gugokha" were divided into form, semantic and climatic elements. As a result, Lee Jung-eom's Cho Young-gwan was able to see the ideal of mountain water and the feeling of being idle in nature as well as the sense of freedom. Sixth, After examining the appearance of the elements and the frequency of the appearance of the landscape, 'water' and 'mountain' were the absolute factors that emphasized the original curved environment at the mouth of Lee Jung-eom. Therefore, there was gugokga can gauge the fresh ideas(神仙思想)and retreat ever(隱居思想). This inherent harmony between the landscape as well as through the mulah any ideas that one with nature and meditation, Confucian tube.

• Journal of Intelligence and Information Systems
• /
• v.20 no.1
• /
• pp.195-211
• /
• 2014

Cloud computing services provide IT resources as services on demand. This is considered a key concept, which will lead a shift from an ownership-based paradigm to a new pay-for-use paradigm, which can reduce the fixed cost for IT resources, and improve flexibility and scalability. As IT services, cloud services have evolved from early similar computing concepts such as network computing, utility computing, server-based computing, and grid computing. So research into cloud computing is highly related to and combined with various relevant computing research areas. To seek promising research issues and topics in cloud computing, it is necessary to understand the research trends in cloud computing more comprehensively. In this study, we collect bibliographic information and citation information for cloud computing related research papers published in major international journals from 1994 to 2012, and analyzes macroscopic trends and network changes to citation relationships among papers and the co-occurrence relationships of key words by utilizing social network analysis measures. Through the analysis, we can identify the relationships and connections among research topics in cloud computing related areas, and highlight new potential research topics. In addition, we visualize dynamic changes of research topics relating to cloud computing using a proposed cloud computing "research trend map." A research trend map visualizes positions of research topics in two-dimensional space. Frequencies of key words (X-axis) and the rates of increase in the degree centrality of key words (Y-axis) are used as the two dimensions of the research trend map. Based on the values of the two dimensions, the two dimensional space of a research map is divided into four areas: maturation, growth, promising, and decline. An area with high keyword frequency, but low rates of increase of degree centrality is defined as a mature technology area; the area where both keyword frequency and the increase rate of degree centrality are high is defined as a growth technology area; the area where the keyword frequency is low, but the rate of increase in the degree centrality is high is defined as a promising technology area; and the area where both keyword frequency and the rate of degree centrality are low is defined as a declining technology area. Based on this method, cloud computing research trend maps make it possible to easily grasp the main research trends in cloud computing, and to explain the evolution of research topics. According to the results of an analysis of citation relationships, research papers on security, distributed processing, and optical networking for cloud computing are on the top based on the page-rank measure. From the analysis of key words in research papers, cloud computing and grid computing showed high centrality in 2009, and key words dealing with main elemental technologies such as data outsourcing, error detection methods, and infrastructure construction showed high centrality in 2010~2011. In 2012, security, virtualization, and resource management showed high centrality. Moreover, it was found that the interest in the technical issues of cloud computing increases gradually. From annual cloud computing research trend maps, it was verified that security is located in the promising area, virtualization has moved from the promising area to the growth area, and grid computing and distributed system has moved to the declining area. The study results indicate that distributed systems and grid computing received a lot of attention as similar computing paradigms in the early stage of cloud computing research. The early stage of cloud computing was a period focused on understanding and investigating cloud computing as an emergent technology, linking to relevant established computing concepts. After the early stage, security and virtualization technologies became main issues in cloud computing, which is reflected in the movement of security and virtualization technologies from the promising area to the growth area in the cloud computing research trend maps. Moreover, this study revealed that current research in cloud computing has rapidly transferred from a focus on technical issues to for a focus on application issues, such as SLAs (Service Level Agreements).

• Journal of Intelligence and Information Systems
• /
• v.20 no.1
• /
• pp.177-193
• /
• 2014

Over the past decade, there has been a rapid diffusion of electronic commerce and a rising number of interconnected networks, resulting in an escalation of security threats and privacy concerns. Electronic commerce has a built-in trade-off between the necessity of providing at least some personal information to consummate an online transaction, and the risk of negative consequences from providing such information. More recently, the frequent disclosure of private information has raised concerns about privacy and its impacts. This has motivated researchers in various fields to explore information privacy issues to address these concerns. Accordingly, the necessity for information privacy policies and technologies for collecting and storing data, and information privacy research in various fields such as medicine, computer science, business, and statistics has increased. The occurrence of various information security accidents have made finding experts in the information security field an important issue. Objective measures for finding such experts are required, as it is currently rather subjective. Based on social network analysis, this paper focused on a framework to evaluate the process of finding experts in the information security field. We collected data from the National Discovery for Science Leaders (NDSL) database, initially collecting about 2000 papers covering the period between 2005 and 2013. Outliers and the data of irrelevant papers were dropped, leaving 784 papers to test the suggested hypotheses. The co-authorship network data for co-author relationship, publisher, affiliation, and so on were analyzed using social network measures including centrality and structural hole. The results of our model estimation are as follows. With the exception of Hypothesis 3, which deals with the relationship between eigenvector centrality and performance, all of our hypotheses were supported. In line with our hypothesis, degree centrality (H1) was supported with its positive influence on the researchers' publishing performance (p<0.001). This finding indicates that as the degree of cooperation increased, the more the publishing performance of researchers increased. In addition, closeness centrality (H2) was also positively associated with researchers' publishing performance (p<0.001), suggesting that, as the efficiency of information acquisition increased, the more the researchers' publishing performance increased. This paper identified the difference in publishing performance among researchers. The analysis can be used to identify core experts and evaluate their performance in the information privacy research field. The co-authorship network for information privacy can aid in understanding the deep relationships among researchers. In addition, extracting characteristics of publishers and affiliations, this paper suggested an understanding of the social network measures and their potential for finding experts in the information privacy field. Social concerns about securing the objectivity of experts have increased, because experts in the information privacy field frequently participate in political consultation, and business education support and evaluation. In terms of practical implications, this research suggests an objective framework for experts in the information privacy field, and is useful for people who are in charge of managing research human resources. This study has some limitations, providing opportunities and suggestions for future research. Presenting the difference in information diffusion according to media and proximity presents difficulties for the generalization of the theory due to the small sample size. Therefore, further studies could consider an increased sample size and media diversity, the difference in information diffusion according to the media type, and information proximity could be explored in more detail. Moreover, previous network research has commonly observed a causal relationship between the independent and dependent variable (Kadushin, 2012). In this study, degree centrality as an independent variable might have causal relationship with performance as a dependent variable. However, in the case of network analysis research, network indices could be computed after the network relationship is created. An annual analysis could help mitigate this limitation.

• Journal of Nutrition and Health
• /
• v.49 no.5
• /
• pp.347-357
• /
• 2016

Purpose: This study investigated bone mineral density and nutritional state according to consumption of milk in Korean postmenopausal women who drink coffee. Methods: Using the 2008~2009 Korean National Health & Nutrition Examination Survey data, a total of 1,373 postmenopausal females aged 50 yrs and over were analyzed after excluding those with diseases related to bone health. According to coffee and/or milk consumption, subjects were divided into four groups: coffee only, both coffee & milk, milk only, and none of the above. All data were processed after application of weighted values and adjustment of age, body mass index, physical activity, drinking, and smoking using a general linear model. For analysis of nutrient intake and bone density, data were additionally adjusted by total energy and calcium intake. Results: The coffee & milk group had more subjects younger than 65 yrs and higher education, urban residents, and higher income than any other group. The coffee only group showed somewhat similar characteristics as the none of the above group, which showed the highest percentage of subjects older than 65 and in a lower education and socio-economic state. Body weight, height, body mass index, and lean mass were the highest in coffee & milk group and lowest in the none of the above group. On the other hand, the milk only group showed the lowest values for body mass index and waist circumference, whereas percent body fat did not show any difference among the groups. The coffee and milk group showed the highest bone mineral density in the total femur and lumbar spine as well as the highest nutritional state and most food group intakes, followed by the milk only group, coffee only group, and none of the above group. In the assessment of osteoporosis based on T-score of bone mineral density, although not significant, the coffee and milk group and milk only group, which showed a better nutritional state, included more subjects with a normal bone density, whereas the none of the above group included more subjects with osteoporosis than any other group. Conclusion: Bone mineral density in postmenopausal women might not be affected by coffee drinking if their diets are accompanied by balanced food and nutrient intake including milk.

• Journal of Intelligence and Information Systems
• /
• v.19 no.2
• /
• pp.141-156
• /
• 2013

Researchers in welfare services of local government in Korea have rather been on isolated issues as disables, childcare, aging phenomenon, etc. (Kang, 2004; Jung et al., 2009). Lately, local officials, yet, realize that they need more comprehensive welfare services for all residents, not just for above-mentioned focused groups. Still cases dealt with focused group approach have been a main research stream due to various reason(Jung et al., 2009; Lee, 2009; Jang, 2011). Social Security Information System is an information system that comprehensively manages 292 welfare benefits provided by 17 ministries and 40 thousand welfare services provided by 230 local authorities in Korea. The purpose of the system is to improve efficiency of social welfare delivery process. The study of local government expenditure has been on the rise over the last few decades after the restarting the local autonomy, but these studies have limitations on data collection. Measurement of a local government's welfare efforts(spending) has been primarily on expenditures or budget for an individual, set aside for welfare. This practice of using monetary value for an individual as a "proxy value" for welfare effort(spending) is based on the assumption that expenditure is directly linked to welfare efforts(Lee et al., 2007). This expenditure/budget approach commonly uses total welfare amount or percentage figure as dependent variables (Wildavsky, 1985; Lee et al., 2007; Kang, 2000). However, current practice of using actual amount being used or percentage figure as a dependent variable may have some limitation; since budget or expenditure is greatly influenced by the total budget of a local government, relying on such monetary value may create inflate or deflate the true "welfare effort" (Jang, 2012). In addition, government budget usually contain a large amount of administrative cost, i.e., salary, for local officials, which is highly unrelated to the actual welfare expenditure (Jang, 2011). This paper used local government welfare service data from the detailed data sets linked to the Social Security Information System. The purpose of this paper is to analyze the factors that affect social welfare spending of 230 local authorities in 2012. The paper applied multiple regression based model to analyze the pooled financial data from the system. Based on the regression analysis, the following factors affecting self-funded welfare spending were identified. In our research model, we use the welfare budget/total budget(%) of a local government as a true measurement for a local government's welfare effort(spending). Doing so, we exclude central government subsidies or support being used for local welfare service. It is because central government welfare support does not truly reflect the welfare efforts(spending) of a local. The dependent variable of this paper is the volume of the welfare spending and the independent variables of the model are comprised of three categories, in terms of socio-demographic perspectives, the local economy and the financial capacity of local government. This paper categorized local authorities into 3 groups, districts, and cities and suburb areas. The model used a dummy variable as the control variable (local political factor). This paper demonstrated that the volume of the welfare spending for the welfare services is commonly influenced by the ratio of welfare budget to total local budget, the population of infants, self-reliance ratio and the level of unemployment factor. Interestingly, the influential factors are different by the size of local government. Analysis of determinants of local government self-welfare spending, we found a significant effect of local Gov. Finance characteristic in degree of the local government's financial independence, financial independence rate, rate of social welfare budget, and regional economic in opening-to-application ratio, and sociology of population in rate of infants. The result means that local authorities should have differentiated welfare strategies according to their conditions and circumstances. There is a meaning that this paper has successfully proven the significant factors influencing welfare spending of local government in Korea.

• Journal of Intelligence and Information Systems
• /
• v.19 no.2
• /
• pp.125-140
• /
• 2013

We call a data set in which the number of records belonging to a certain class far outnumbers the number of records belonging to the other class, 'imbalanced data set'. Most of the classification techniques perform poorly on imbalanced data sets. When we evaluate the performance of a certain classification technique, we need to measure not only 'accuracy' but also 'sensitivity' and 'specificity'. In a customer churn prediction problem, 'retention' records account for the majority class, and 'churn' records account for the minority class. Sensitivity measures the proportion of actual retentions which are correctly identified as such. Specificity measures the proportion of churns which are correctly identified as such. The poor performance of the classification techniques on imbalanced data sets is due to the low value of specificity. Many previous researches on imbalanced data sets employed 'oversampling' technique where members of the minority class are sampled more than those of the majority class in order to make a relatively balanced data set. When a classification model is constructed using this oversampled balanced data set, specificity can be improved but sensitivity will be decreased. In this research, we developed a hybrid model of support vector machine (SVM), artificial neural network (ANN) and decision tree, that improves specificity while maintaining sensitivity. We named this hybrid model 'hybrid SVM model.' The process of construction and prediction of our hybrid SVM model is as follows. By oversampling from the original imbalanced data set, a balanced data set is prepared. SVM_I model and ANN_I model are constructed using the imbalanced data set, and SVM_B model is constructed using the balanced data set. SVM_I model is superior in sensitivity and SVM_B model is superior in specificity. For a record on which both SVM_I model and SVM_B model make the same prediction, that prediction becomes the final solution. If they make different prediction, the final solution is determined by the discrimination rules obtained by ANN and decision tree. For a record on which SVM_I model and SVM_B model make different predictions, a decision tree model is constructed using ANN_I output value as input and actual retention or churn as target. We obtained the following two discrimination rules: 'IF ANN_I output value <0.285, THEN Final Solution = Retention' and 'IF ANN_I output value ${\geq}0.285$, THEN Final Solution = Churn.' The threshold 0.285 is the value optimized for the data used in this research. The result we present in this research is the structure or framework of our hybrid SVM model, not a specific threshold value such as 0.285. Therefore, the threshold value in the above discrimination rules can be changed to any value depending on the data. In order to evaluate the performance of our hybrid SVM model, we used the 'churn data set' in UCI Machine Learning Repository, that consists of 85% retention customers and 15% churn customers. Accuracy of the hybrid SVM model is 91.08% that is better than that of SVM_I model or SVM_B model. The points worth noticing here are its sensitivity, 95.02%, and specificity, 69.24%. The sensitivity of SVM_I model is 94.65%, and the specificity of SVM_B model is 67.00%. Therefore the hybrid SVM model developed in this research improves the specificity of SVM_B model while maintaining the sensitivity of SVM_I model.

• Journal of Intelligence and Information Systems
• /
• v.25 no.4
• /
• pp.141-154
• /
• 2019

Rapid growth of internet technology and social media is progressing. Data mining technology has evolved to enable unstructured document representations in a variety of applications. Sentiment analysis is an important technology that can distinguish poor or high-quality content through text data of products, and it has proliferated during text mining. Sentiment analysis mainly analyzes people's opinions in text data by assigning predefined data categories as positive and negative. This has been studied in various directions in terms of accuracy from simple rule-based to dictionary-based approaches using predefined labels. In fact, sentiment analysis is one of the most active researches in natural language processing and is widely studied in text mining. When real online reviews aren't available for others, it's not only easy to openly collect information, but it also affects your business. In marketing, real-world information from customers is gathered on websites, not surveys. Depending on whether the website's posts are positive or negative, the customer response is reflected in the sales and tries to identify the information. However, many reviews on a website are not always good, and difficult to identify. The earlier studies in this research area used the reviews data of the Amazon.com shopping mal, but the research data used in the recent studies uses the data for stock market trends, blogs, news articles, weather forecasts, IMDB, and facebook etc. However, the lack of accuracy is recognized because sentiment calculations are changed according to the subject, paragraph, sentiment lexicon direction, and sentence strength. This study aims to classify the polarity analysis of sentiment analysis into positive and negative categories and increase the prediction accuracy of the polarity analysis using the pretrained IMDB review data set. First, the text classification algorithm related to sentiment analysis adopts the popular machine learning algorithms such as NB (naive bayes), SVM (support vector machines), XGboost, RF (random forests), and Gradient Boost as comparative models. Second, deep learning has demonstrated discriminative features that can extract complex features of data. Representative algorithms are CNN (convolution neural networks), RNN (recurrent neural networks), LSTM (long-short term memory). CNN can be used similarly to BoW when processing a sentence in vector format, but does not consider sequential data attributes. RNN can handle well in order because it takes into account the time information of the data, but there is a long-term dependency on memory. To solve the problem of long-term dependence, LSTM is used. For the comparison, CNN and LSTM were chosen as simple deep learning models. In addition to classical machine learning algorithms, CNN, LSTM, and the integrated models were analyzed. Although there are many parameters for the algorithms, we examined the relationship between numerical value and precision to find the optimal combination. And, we tried to figure out how the models work well for sentiment analysis and how these models work. This study proposes integrated CNN and LSTM algorithms to extract the positive and negative features of text analysis. The reasons for mixing these two algorithms are as follows. CNN can extract features for the classification automatically by applying convolution layer and massively parallel processing. LSTM is not capable of highly parallel processing. Like faucets, the LSTM has input, output, and forget gates that can be moved and controlled at a desired time. These gates have the advantage of placing memory blocks on hidden nodes. The memory block of the LSTM may not store all the data, but it can solve the CNN's long-term dependency problem. Furthermore, when LSTM is used in CNN's pooling layer, it has an end-to-end structure, so that spatial and temporal features can be designed simultaneously. In combination with CNN-LSTM, 90.33% accuracy was measured. This is slower than CNN, but faster than LSTM. The presented model was more accurate than other models. In addition, each word embedding layer can be improved when training the kernel step by step. CNN-LSTM can improve the weakness of each model, and there is an advantage of improving the learning by layer using the end-to-end structure of LSTM. Based on these reasons, this study tries to enhance the classification accuracy of movie reviews using the integrated CNN-LSTM model.

• Journal of Intelligence and Information Systems
• /
• v.26 no.2
• /
• pp.131-145
• /
• 2020

In line with the trend of industrial innovation, IoT technology utilized in a variety of fields is emerging as a key element in creation of new business models and the provision of user-friendly services through the combination of big data. The accumulated data from devices with the Internet-of-Things (IoT) is being used in many ways to build a convenience-based smart system as it can provide customized intelligent systems through user environment and pattern analysis. Recently, it has been applied to innovation in the public domain and has been using it for smart city and smart transportation, such as solving traffic and crime problems using CCTV. In particular, it is necessary to comprehensively consider the easiness of securing real-time service data and the stability of security when planning underground services or establishing movement amount control information system to enhance citizens' or commuters' convenience in circumstances with the congestion of public transportation such as subways, urban railways, etc. However, previous studies that utilize image data have limitations in reducing the performance of object detection under private issue and abnormal conditions. The IoT device-based sensor data used in this study is free from private issue because it does not require identification for individuals, and can be effectively utilized to build intelligent public services for unspecified people. Especially, sensor data stored by the IoT device need not be identified to an individual, and can be effectively utilized for constructing intelligent public services for many and unspecified people as data free form private issue. We utilize the IoT-based infrared sensor devices for an intelligent pedestrian tracking system in metro service which many people use on a daily basis and temperature data measured by sensors are therein transmitted in real time. The experimental environment for collecting data detected in real time from sensors was established for the equally-spaced midpoints of 4×4 upper parts in the ceiling of subway entrances where the actual movement amount of passengers is high, and it measured the temperature change for objects entering and leaving the detection spots. The measured data have gone through a preprocessing in which the reference values for 16 different areas are set and the difference values between the temperatures in 16 distinct areas and their reference values per unit of time are calculated. This corresponds to the methodology that maximizes movement within the detection area. In addition, the size of the data was increased by 10 times in order to more sensitively reflect the difference in temperature by area. For example, if the temperature data collected from the sensor at a given time were 28.5℃, the data analysis was conducted by changing the value to 285. As above, the data collected from sensors have the characteristics of time series data and image data with 4×4 resolution. Reflecting the characteristics of the measured, preprocessed data, we finally propose a hybrid algorithm that combines CNN in superior performance for image classification and LSTM, especially suitable for analyzing time series data, as referred to CNN-LSTM (Convolutional Neural Network-Long Short Term Memory). In the study, the CNN-LSTM algorithm is used to predict the number of passing persons in one of 4×4 detection areas. We verified the validation of the proposed model by taking performance comparison with other artificial intelligence algorithms such as Multi-Layer Perceptron (MLP), Long Short Term Memory (LSTM) and RNN-LSTM (Recurrent Neural Network-Long Short Term Memory). As a result of the experiment, proposed CNN-LSTM hybrid model compared to MLP, LSTM and RNN-LSTM has the best predictive performance. By utilizing the proposed devices and models, it is expected various metro services will be provided with no illegal issue about the personal information such as real-time monitoring of public transport facilities and emergency situation response services on the basis of congestion. However, the data have been collected by selecting one side of the entrances as the subject of analysis, and the data collected for a short period of time have been applied to the prediction. There exists the limitation that the verification of application in other environments needs to be carried out. In the future, it is expected that more reliability will be provided for the proposed model if experimental data is sufficiently collected in various environments or if learning data is further configured by measuring data in other sensors.

• Journal of Intelligence and Information Systems
• /
• v.26 no.2
• /
• pp.1-25
• /
• 2020

In this paper, we suggest an application system architecture which provides accurate, fast and efficient automatic gasometer reading function. The system captures gasometer image using mobile device camera, transmits the image to a cloud server on top of private LTE network, and analyzes the image to extract character information of device ID and gas usage amount by selective optical character recognition based on deep learning technology. In general, there are many types of character in an image and optical character recognition technology extracts all character information in an image. But some applications need to ignore non-of-interest types of character and only have to focus on some specific types of characters. For an example of the application, automatic gasometer reading system only need to extract device ID and gas usage amount character information from gasometer images to send bill to users. Non-of-interest character strings, such as device type, manufacturer, manufacturing date, specification and etc., are not valuable information to the application. Thus, the application have to analyze point of interest region and specific types of characters to extract valuable information only. We adopted CNN (Convolutional Neural Network) based object detection and CRNN (Convolutional Recurrent Neural Network) technology for selective optical character recognition which only analyze point of interest region for selective character information extraction. We build up 3 neural networks for the application system. The first is a convolutional neural network which detects point of interest region of gas usage amount and device ID information character strings, the second is another convolutional neural network which transforms spatial information of point of interest region to spatial sequential feature vectors, and the third is bi-directional long short term memory network which converts spatial sequential information to character strings using time-series analysis mapping from feature vectors to character strings. In this research, point of interest character strings are device ID and gas usage amount. Device ID consists of 12 arabic character strings and gas usage amount consists of 4 ~ 5 arabic character strings. All system components are implemented in Amazon Web Service Cloud with Intel Zeon E5-2686 v4 CPU and NVidia TESLA V100 GPU. The system architecture adopts master-lave processing structure for efficient and fast parallel processing coping with about 700,000 requests per day. Mobile device captures gasometer image and transmits to master process in AWS cloud. Master process runs on Intel Zeon CPU and pushes reading request from mobile device to an input queue with FIFO (First In First Out) structure. Slave process consists of 3 types of deep neural networks which conduct character recognition process and runs on NVidia GPU module. Slave process is always polling the input queue to get recognition request. If there are some requests from master process in the input queue, slave process converts the image in the input queue to device ID character string, gas usage amount character string and position information of the strings, returns the information to output queue, and switch to idle mode to poll the input queue. Master process gets final information form the output queue and delivers the information to the mobile device. We used total 27,120 gasometer images for training, validation and testing of 3 types of deep neural network. 22,985 images were used for training and validation, 4,135 images were used for testing. We randomly splitted 22,985 images with 8:2 ratio for training and validation respectively for each training epoch. 4,135 test image were categorized into 5 types (Normal, noise, reflex, scale and slant). Normal data is clean image data, noise means image with noise signal, relfex means image with light reflection in gasometer region, scale means images with small object size due to long-distance capturing and slant means images which is not horizontally flat. Final character string recognition accuracies for device ID and gas usage amount of normal data are 0.960 and 0.864 respectively.

• Journal of Intelligence and Information Systems
• /
• v.26 no.2
• /
• pp.79-104
• /
• 2020

Recently, as deep learning has attracted attention, the use of deep learning is being considered as a method for solving problems in various fields. In particular, deep learning is known to have excellent performance when applied to applying unstructured data such as text, sound and images, and many studies have proven its effectiveness. Owing to the remarkable development of text and image deep learning technology, interests in image captioning technology and its application is rapidly increasing. Image captioning is a technique that automatically generates relevant captions for a given image by handling both image comprehension and text generation simultaneously. In spite of the high entry barrier of image captioning that analysts should be able to process both image and text data, image captioning has established itself as one of the key fields in the A.I. research owing to its various applicability. In addition, many researches have been conducted to improve the performance of image captioning in various aspects. Recent researches attempt to create advanced captions that can not only describe an image accurately, but also convey the information contained in the image more sophisticatedly. Despite many recent efforts to improve the performance of image captioning, it is difficult to find any researches to interpret images from the perspective of domain experts in each field not from the perspective of the general public. Even for the same image, the part of interests may differ according to the professional field of the person who has encountered the image. Moreover, the way of interpreting and expressing the image also differs according to the level of expertise. The public tends to recognize the image from a holistic and general perspective, that is, from the perspective of identifying the image's constituent objects and their relationships. On the contrary, the domain experts tend to recognize the image by focusing on some specific elements necessary to interpret the given image based on their expertise. It implies that meaningful parts of an image are mutually different depending on viewers' perspective even for the same image. So, image captioning needs to implement this phenomenon. Therefore, in this study, we propose a method to generate captions specialized in each domain for the image by utilizing the expertise of experts in the corresponding domain. Specifically, after performing pre-training on a large amount of general data, the expertise in the field is transplanted through transfer-learning with a small amount of expertise data. However, simple adaption of transfer learning using expertise data may invoke another type of problems. Simultaneous learning with captions of various characteristics may invoke so-called 'inter-observation interference' problem, which make it difficult to perform pure learning of each characteristic point of view. For learning with vast amount of data, most of this interference is self-purified and has little impact on learning results. On the contrary, in the case of fine-tuning where learning is performed on a small amount of data, the impact of such interference on learning can be relatively large. To solve this problem, therefore, we propose a novel 'Character-Independent Transfer-learning' that performs transfer learning independently for each character. In order to confirm the feasibility of the proposed methodology, we performed experiments utilizing the results of pre-training on MSCOCO dataset which is comprised of 120,000 images and about 600,000 general captions. Additionally, according to the advice of an art therapist, about 300 pairs of 'image / expertise captions' were created, and the data was used for the experiments of expertise transplantation. As a result of the experiment, it was confirmed that the caption generated according to the proposed methodology generates captions from the perspective of implanted expertise whereas the caption generated through learning on general data contains a number of contents irrelevant to expertise interpretation. In this paper, we propose a novel approach of specialized image interpretation. To achieve this goal, we present a method to use transfer learning and generate captions specialized in the specific domain. In the future, by applying the proposed methodology to expertise transplant in various fields, we expected that many researches will be actively conducted to solve the problem of lack of expertise data and to improve performance of image captioning.