• Smart Media Journal
• /
• v.11 no.10
• /
• pp.65-75
• /
• 2022

Deep learning is used as a creative tool that could overcome the limitations of existing analysis models and generate various types of results such as text, image, and music. In this paper, we propose a method necessary to preprocess audio data using the Niko's MIDI Pack sound source file as a data set and to generate music using Bi-LSTM. Based on the generated root note, the hidden layers are composed of multi-layers to create a new note suitable for the musical composition, and an attention mechanism is applied to the output gate of the decoder to apply the weight of the factors that affect the data input from the encoder. Setting variables such as loss function and optimization method are applied as parameters for improving the LSTM model. The proposed model is a multi-channel Bi-LSTM with attention that applies notes pitch generated from separating treble clef and bass clef, length of notes, rests, length of rests, and chords to improve the efficiency and prediction of MIDI deep learning process. The results of the learning generate a sound that matches the development of music scale distinct from noise, and we are aiming to contribute to generating a harmonistic stable music.

• Journal of KIISE
• /
• v.43 no.10
• /
• pp.1085-1093
• /
• 2016

Community-based Question Answering system is a system which provides answers for each question from the documents uploaded on web communities. In order to enhance the capacity of question analysis, former methods have developed specific rules suitable for a target region or have applied machine learning to partial processes. However, these methods incur an excessive cost for expanding fields or lead to cases in which system is overfitted for a specific field. This paper proposes a multiple machine learning method which automates the overall process by adapting appropriate machine learning in each procedure for efficient processing of community-based Question Answering system. This system can be divided into question analysis part and answer selection part. The question analysis part consists of the question focus extractor, which analyzes the focused phrases in questions and uses conditional random fields, and the question type classifier, which classifies topics of questions and uses support vector machine. In the answer selection part, the we trains weights that are used by the similarity estimation models through an artificial neural network. Also these are a number of cases in which the results of morphological analysis are not reliable for the data uploaded on web communities. Therefore, we suggest a method that minimizes the impact of morphological analysis by using character features in the stage of question analysis. The proposed system outperforms the former system by showing a Mean Average Precision criteria of 0.765 and R-Precision criteria of 0.872.

• Journal of Information Management
• /
• v.25 no.1
• /
• pp.24-46
• /
• 1994

In this paper I will discuss definitions of hypermedia, multimedia and hypertext. Hypertext is the grouping of relevant information in the form of nodes. These nodes are then connected together through links. In the case of hypertext the nodes contain text or graphics. Multimedia is the combining of different media types for example sound, animation, text, graphics and video for the presentation of information by making use of computers. Hypermedia can be viewed as an extension of hypertext and multimedia. It is based on the concept of hypertext that uses nodes and links in the structuring of information in the system. In this case the nodes consist of an the different data types that are mentioned in the multimedia definition above. The 'node-and-link' concept is used in organisation of the information in hypermedia systems. The 'book' metaphor is an example of the way these systems are implemented. This concept is explained and a few advantages and disadvantages of making use of hypermedia systems are discussed. A new approach for the development of hypermedia systems, namely the knowledge-based approach is now looked into. Joel Peing-Ling Loo proposed this approach because he thought that it is the most effective way for handling this kind of technology. A semantic-based hypermedia model is developed in this approach to formulate solutions for the restrictions in presenting information authoring, maintenance and retrieval. The knowledge-based presentation of information includes the use of conventional data structures. These data structures make use of frames(objects), slots and the inheritance theory that is also used in expert systems. Relations develop between the different objects as these objects are included in the database. Relations can also exist between frames by means of attributes that belong to the frames.

• Journal of Intelligence and Information Systems
• /
• v.27 no.3
• /
• pp.157-173
• /
• 2021

With the development of information technology, the amount of available information increases daily. However, having access to so much information makes it difficult for users to easily find the information they seek. Users want a visualized system that reduces information retrieval and learning time, saving them from personally reading and judging all available information. As a result, recommendation systems are an increasingly important technologies that are essential to the business. Collaborative filtering is used in various fields with excellent performance because recommendations are made based on similar user interests and preferences. However, limitations do exist. Sparsity occurs when user-item preference information is insufficient, and is the main limitation of collaborative filtering. The evaluation value of the user item matrix may be distorted by the data depending on the popularity of the product, or there may be new users who have not yet evaluated the value. The lack of historical data to identify consumer preferences is referred to as data sparsity, and various methods have been studied to address these problems. However, most attempts to solve the sparsity problem are not optimal because they can only be applied when additional data such as users' personal information, social networks, or characteristics of items are included. Another problem is that real-world score data are mostly biased to high scores, resulting in severe imbalances. One cause of this imbalance distribution is the purchasing bias, in which only users with high product ratings purchase products, so those with low ratings are less likely to purchase products and thus do not leave negative product reviews. Due to these characteristics, unlike most users' actual preferences, reviews by users who purchase products are more likely to be positive. Therefore, the actual rating data is over-learned in many classes with high incidence due to its biased characteristics, distorting the market. Applying collaborative filtering to these imbalanced data leads to poor recommendation performance due to excessive learning of biased classes. Traditional oversampling techniques to address this problem are likely to cause overfitting because they repeat the same data, which acts as noise in learning, reducing recommendation performance. In addition, pre-processing methods for most existing data imbalance problems are designed and used for binary classes. Binary class imbalance techniques are difficult to apply to multi-class problems because they cannot model multi-class problems, such as objects at cross-class boundaries or objects overlapping multiple classes. To solve this problem, research has been conducted to convert and apply multi-class problems to binary class problems. However, simplification of multi-class problems can cause potential classification errors when combined with the results of classifiers learned from other sub-problems, resulting in loss of important information about relationships beyond the selected items. Therefore, it is necessary to develop more effective methods to address multi-class imbalance problems. We propose a collaborative filtering model using CGAN to generate realistic virtual data to populate the empty user-item matrix. Conditional vector y identify distributions for minority classes and generate data reflecting their characteristics. Collaborative filtering then maximizes the performance of the recommendation system via hyperparameter tuning. This process should improve the accuracy of the model by addressing the sparsity problem of collaborative filtering implementations while mitigating data imbalances arising from real data. Our model has superior recommendation performance over existing oversampling techniques and existing real-world data with data sparsity. SMOTE, Borderline SMOTE, SVM-SMOTE, ADASYN, and GAN were used as comparative models and we demonstrate the highest prediction accuracy on the RMSE and MAE evaluation scales. Through this study, oversampling based on deep learning will be able to further refine the performance of recommendation systems using actual data and be used to build business recommendation systems.

• Journal of the Korean earth science society
• /
• v.40 no.5
• /
• pp.502-517
• /
• 2019

Global warming and rapid climate change have long affected the characteristics of typhoons in the Northwest Pacific, which has induced increasing devastating disasters along the coastal regions of the Korean peninsula. Synthetic Aperature Radar (SAR), as one of the microwave sensors, makes it possible to produce high-resolution sea surface wind field around the typhoon under cloudy atmospheric conditions, which has been impossible to obtain the winds from satellite optical and infrared sensors. The Geophysical Model Functions (GMFs) for sea surface wind retrieval from SAR data requires the input of wind direction, which should be based on the accurate estimation of the center of the typhoon. This study estimated the typhoon centers using Sentinel-1A images to improve the problem of typhoon center detection method and to reflect it in retrieving the sea surface wind. The results were validated by comparing with the typhoon best track data provided by the Korea Meteorological Administration (KMA) and Japan Meteorological Agency (JMA), and also by using infrared images of Himawari-8 satellite. The initial center position of the typhoon was determined by using VH polarization, thereby reducing the possibility of error. The detected center showed a difference of 23.76 km on average with the best track data of the four typhoons provided by the KMA and JMA. Compared to the typhoon center estimated by Himawari-8 satellite, the results showed an average spatial variation of 11.80 km except one typhoon located near land with a large difference of 58.73 km. This result suggests that high-resolution SAR images can be used to estimate the center and retrieve sea surface wind around typhoons.

• Journal of the Korean earth science society
• /
• v.39 no.1
• /
• pp.53-66
• /
• 2018

The information of surface reflectance ($R_{sfc}$) is important for the heat balance and the environmental/climate monitoring. The $R_{sfc}$ sensitivity to error-induced variables for the Geostationary Environment Monitoring Spectrometer (GEMS) retrieval from geostationary-orbit satellite observations at 300-500 nm was investigated, utilizing polar-orbit satellite data of the MODerate resolution Imaging Spectroradiometer (MODIS) and Ozone Mapping Instrument (OMI), and the radiative transfer model (RTM) experiment. The variables in this study can be cloud, Rayleigh-scattering, aerosol, ozone and surface type. The cloud detection in high-resolution MODIS pixels ($1km{\times}1km$) was compared with that in GEMS-scale pixels ($8km{\times}7km$). The GEMS detection was consistent (~79%) with the MODIS result. However, the detection probability in partially-cloudy (${\leq}40%$) GEMS pixels decreased due to other effects (i.e., aerosol and surface type). The Rayleigh-scattering effect in RGB images was noticeable over ocean, based on the RTM calculation. The reflectance at top of atmosphere ($R_{toa}$) increased with aerosol amounts in case of $R_{sfc}$<0.2, but decreased in $R_{sfc}{\geq}0.2$. The $R_{sfc}$ errors due to the aerosol increased with wavelength in the UV, but were constant or slightly decreased in the visible. The ozone absorption was most sensitive at 328 nm in the UV region (328-354 nm). The $R_{sfc}$ error was +0.1 because of negative total ozone anomaly (-100 DU) under the condition of $R_{sfc}=0.15$. This study can be useful to estimate $R_{sfc}$ uncertainties in the GEMS retrieval.

• The KIPS Transactions:PartB
• /
• v.19B no.3
• /
• pp.189-194
• /
• 2012

In this paper, we propose a query expansion method based on word graphs using pseudo-relevant and pseudo non-relevant documents to achieve performance improvement in information retrieval. The initially retrieved documents are classified into a core cluster when a document includes core query terms extracted by query term combinations and the degree of query term proximity. Otherwise, documents are classified into a non-core cluster. The documents that belong to a core query cluster can be seen as pseudo-relevant documents, and the documents that belong to a non-core cluster can be seen as pseudo non-relevant documents. Each cluster is represented as a graph which has nodes and edges. Each node represents a term and each edge represents proximity between the term and a query term. The term weight is calculated by subtracting the term weight in the non-core cluster graph from the term weight in the core cluster graph. It means that a term with a high weight in a non-core cluster graph should not be considered as an expanded term. Expansion terms are selected according to the term weights. Experimental results on TREC WT10g test collection show that the proposed method achieves 9.4% improvement over the language model in mean average precision.

• Journal of Information Management
• /
• v.41 no.4
• /
• pp.227-249
• /
• 2010

Scholarly information has increased tremendously according to the development of IT, especially the Internet. However, simultaneously, people have to spend more time and exert more effort because of information overload. There have been many research efforts in the field of expert systems, data mining, and information retrieval, concerning a system that recommends user-expected information items through presumption. Recently, the hybrid system combining a content-based recommendation system and collaborative filtering or combining recommendation systems in other domains has been developed. In this paper we resolved the problem of the current recommendation system and suggested a new system combining collaborative filtering and Naive Bayes Classification. In this way, we resolved the over-specialization problem through collaborative filtering and lack of assessment information or recommendation of new contents through Naive Bayes Classification. For verification, we applied the new model in NDSL's paper service of KISTI, especially papers from journals about Sitology and Electronics, and witnessed high satisfaction from 4 experimental participants.

• Journal of KIISE:Information Networking
• /
• v.29 no.3
• /
• pp.286-298
• /
• 2002

The Remote Procedure Call(RPC) has been traditionally used for Inter Process Communication(IPC) among precesses in distributed computing environment. As distributed applications have been complicated more and more, the Mobile Agent paradigm for IPC is emerged. Because there are some paradigms for IPC, researches to evaluate and compare the performance of each paradigm are issued recently. But the performance models used in the previous research did not reflect real distributed computing environment correctly, because they did not consider the evacuation elements for providing security services. Since real distributed environment is open, it is very vulnerable to a variety of attacks. In order to execute applications securely in distributed computing environment, security services which protect applications and information against the attacks must be considered. In this paper, we evaluate and compare the performance of the Remote Procedure Call with that of the Mobile Agent in IPC paradigms. We examine security services to execute applications securely, and propose new performance models considering those services. We design performance models, which describe information retrieval system through N database services, using Petri Net. We compare the performance of two paradigms by assigning numerical values to parameters and measuring the execution time of two paradigms. In this paper, the comparison of two performance models with security services for secure communication shows the results that the execution time of the Remote Procedure Call performance model is sharply increased because of many communications with the high cryptography mechanism between hosts, and that the execution time of the Mobile Agent model is gradually increased because the Mobile Agent paradigm can reduce the quantity of the communications between hosts.

• Journal of Intelligence and Information Systems
• /
• v.24 no.3
• /
• pp.1-19
• /
• 2018

Large amount of data is now available for research and business sectors to extract knowledge from it. This data can be in the form of unstructured data such as audio, text, and image data and can be analyzed by deep learning methodology. Deep learning is now widely used for various estimation, classification, and prediction problems. Especially, fashion business adopts deep learning techniques for apparel recognition, apparel search and retrieval engine, and automatic product recommendation. The core model of these applications is the image classification using Convolutional Neural Networks (CNN). CNN is made up of neurons which learn parameters such as weights while inputs come through and reach outputs. CNN has layer structure which is best suited for image classification as it is comprised of convolutional layer for generating feature maps, pooling layer for reducing the dimensionality of feature maps, and fully-connected layer for classifying the extracted features. However, most of the classification models have been trained using online product image, which is taken under controlled situation such as apparel image itself or professional model wearing apparel. This image may not be an effective way to train the classification model considering the situation when one might want to classify street fashion image or walking image, which is taken in uncontrolled situation and involves people's movement and unexpected pose. Therefore, we propose to train the model with runway apparel image dataset which captures mobility. This will allow the classification model to be trained with far more variable data and enhance the adaptation with diverse query image. To achieve both convergence and generalization of the model, we apply Transfer Learning on our training network. As Transfer Learning in CNN is composed of pre-training and fine-tuning stages, we divide the training step into two. First, we pre-train our architecture with large-scale dataset, ImageNet dataset, which consists of 1.2 million images with 1000 categories including animals, plants, activities, materials, instrumentations, scenes, and foods. We use GoogLeNet for our main architecture as it has achieved great accuracy with efficiency in ImageNet Large Scale Visual Recognition Challenge (ILSVRC). Second, we fine-tune the network with our own runway image dataset. For the runway image dataset, we could not find any previously and publicly made dataset, so we collect the dataset from Google Image Search attaining 2426 images of 32 major fashion brands including Anna Molinari, Balenciaga, Balmain, Brioni, Burberry, Celine, Chanel, Chloe, Christian Dior, Cividini, Dolce and Gabbana, Emilio Pucci, Ermenegildo, Fendi, Giuliana Teso, Gucci, Issey Miyake, Kenzo, Leonard, Louis Vuitton, Marc Jacobs, Marni, Max Mara, Missoni, Moschino, Ralph Lauren, Roberto Cavalli, Sonia Rykiel, Stella McCartney, Valentino, Versace, and Yve Saint Laurent. We perform 10-folded experiments to consider the random generation of training data, and our proposed model has achieved accuracy of 67.2% on final test. Our research suggests several advantages over previous related studies as to our best knowledge, there haven't been any previous studies which trained the network for apparel image classification based on runway image dataset. We suggest the idea of training model with image capturing all the possible postures, which is denoted as mobility, by using our own runway apparel image dataset. Moreover, by applying Transfer Learning and using checkpoint and parameters provided by Tensorflow Slim, we could save time spent on training the classification model as taking 6 minutes per experiment to train the classifier. This model can be used in many business applications where the query image can be runway image, product image, or street fashion image. To be specific, runway query image can be used for mobile application service during fashion week to facilitate brand search, street style query image can be classified during fashion editorial task to classify and label the brand or style, and website query image can be processed by e-commerce multi-complex service providing item information or recommending similar item.