• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.8
• /
• pp.341-354
• /
• 2023

The title of a document is the brief summarization of the document. Readers can easily understand a document if we provide them with its title in their preferred styles and the languages. In this research, we propose a cross-lingual and style-based title generation model using multiple adapters. To train the model, we need a parallel corpus in several languages with different styles. It is quite difficult to construct this kind of parallel corpus; however, a monolingual title generation corpus of the same style can be built easily. Therefore, we apply a zero-shot strategy to generate a title in a different language and with a different style for an input document. A baseline model is Transformer consisting of an encoder and a decoder, pre-trained by several languages. The model is then equipped with multiple adapters for translation, languages, and styles. After the model learns a translation task from parallel corpus, it learns a title generation task from monolingual title generation corpus. When training the model with a task, we only activate an adapter that corresponds to the task. When generating a cross-lingual and style-based title, we only activate adapters that correspond to a target language and a target style. An experimental result shows that our proposed model is only as good as a pipeline model that first translates into a target language and then generates a title. There have been significant changes in natural language generation due to the emergence of large-scale language models. However, research to improve the performance of natural language generation using limited resources and limited data needs to continue. In this regard, this study seeks to explore the significance of such research.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.141-143
• /
• 2022

This paper presented a new technology to identify traffic violations of motorbike riders by detecting the traffic signal using You Only Look Once (YOLO) object detection. The hardware module that is mounted on the front of the motorbike consists of Raspberry Pi with a camera to run the YOLO object detection, a GPS module to acquire the motorcycle's coordinate, and a LoRa communication module to send the data to a cloud DB. The main goal of the software is to determine whether a motorbike has violated a traffic signal. This paper proposes a function to recognize the red traffic signal colour with its movement inside the camera angle and determine that the traffic signal violation happens if the traffic signal is moving to the right direction (the rider turns left) or moving to the top direction (the riders goes straight). Furthermore, if a motorbike rider is violated the signal, the rider's personal information (name, mobile phone number, etc), the snapshot of the violation situation, rider's location, and date/time will be sent to a cloud DB. The violation information will be delivered to the driver's smartphone as a push notification and the local police station to be used for issuing violation tickets, which is expected to prevent motorbike riders from violating traffic signals.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.10a
• /
• pp.433-435
• /
• 2022

In this paper, we introduce a computing resource standard metadata schema design plan for registering, retrieving, and managing computing resources used for research data analysis and utilization in the Korea Research Data Commons(KRDC). KRDC is a joint utilization system of research data and computing resources to maximize the sharing and utilization of research data. Computing resources refer to all resources in the computing environment, such as analysis infrastructure and analysis software, necessary to analyze and utilize research data used in the entire research process. The standard metadata schema for KRDC computing resource management is designed by considering common attributes for computing resource management and other attributes according to each computing resource feature. The standard metadata schema for computing resource management consists of a computing resource metadata schema and a computing resource provider metadata schema. In addition, the metadata schema of computing resources and providers was designed as a service schema and a system schema group according to their characteristics. The standard metadata schema designed in this paper is used for computing resource registration, retrieval, management, and workflow services for computing resource providers and computing resource users through the KRDC web service, and is designed in a scalable form for various computing resource links.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.4
• /
• pp.149-158
• /
• 2023

As data utilization is becoming more important recently, the importance of data centers is also increasing. However, the data center is a problem in terms of environment and economy because it is a massive power-consuming facility that runs 24 hours a day. Recently, studies using deep learning techniques to reduce power used in data centers or servers or predict traffic have been conducted from various perspectives. However, the amount of traffic data processed by the server is anomalous, which makes it difficult to manage the server. In addition, many studies on dynamic server management techniques are still required. Therefore, in this paper, we propose a dynamic server management technique based on Long-Term Short Memory (LSTM), which is robust to time series data prediction. The proposed model allows servers to be managed more reliably and efficiently in the field environment than before, and reduces power used by servers more effectively. For verification of the proposed model, we collect transmission and reception traffic data from six of Wikipedia's data centers, and then analyze and experiment with statistical-based analysis on the relationship of each traffic data. Experimental results show that the proposed model is helpful for reliably and efficiently running servers.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.4
• /
• pp.159-172
• /
• 2023

In general, social problem-solving research aims to create important social value by offering meaningful answers to various social pending issues using scientific technologies. Not surprisingly, however, although numerous and extensive research attempts have been made to alleviate the social problems and issues in nation-wide, we still have many important social challenges and works to be done. In order to facilitate the entire process of the social problem-solving research and maximize its efficacy, it is vital to clearly identify and grasp the important and pressing problems to be focused upon. It is understandable for the problem discovery step to be drastically improved if current social issues can be automatically identified from existing R&D resources such as technical reports and articles. This paper introduces a comprehensive dataset which is essential to build a machine learning model for automatically detecting the social problems and solutions in various national research reports. Initially, we collected a total of 700 research reports regarding social problems and issues. Through intensive annotation process, we built totally 24,022 sentences each of which possesses its own category or label closely related to social problem-solving such as problems, purposes, solutions, effects and so on. Furthermore, we implemented four sentence classification models based on various neural language models and conducted a series of performance experiments using our dataset. As a result of the experiment, the model fine-tuned to the KLUE-BERT pre-trained language model showed the best performance with an accuracy of 75.853% and an F1 score of 63.503%.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.4
• /
• pp.173-178
• /
• 2023

As Korean literature spreads around the world, its position in the overseas publishing market has become important. As demand in the overseas publishing market continues to grow, it is essential to predict future book sales and analyze the characteristics of books that have been highly favored by overseas readers in the past. In this study, we proposed ensemble learning based prediction model and analyzed characteristics of the cumulative sales of more than 5,000 copies classified as good sellers published overseas over the past 5 years. We applied the five ensemble learning models, i.e., XGBoost, Gradient Boosting, Adaboost, LightGBM, and Random Forest, and compared them with other machine learning algorithms, i.e., Support Vector Machine, Logistic Regression, and Deep Learning. Our experimental results showed that the ensemble algorithm outperforms other approaches in troubleshooting imbalanced data. In particular, the LightGBM model obtained an AUC value of 99.86% which is the best prediction performance. Among the features used for prediction, the most important feature is the author's number of overseas publications, and the second important feature is publication in countries with the largest publication market size. The number of evaluation participants is also an important feature. In addition, text mining was performed on the four book reviews that sold the most among good-selling books. Many reviews were interested in stories, characters, and writers and it seems that support for translation is needed as many of the keywords of "translation" appear in low-rated reviews.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.4
• /
• pp.179-188
• /
• 2023

Recently, fake news disguises the form of news content and appears whenever important events occur, causing social confusion. Accordingly, artificial intelligence technology is used as a research to detect fake news. Fake news detection approaches such as automatically recognizing and blocking fake news through natural language processing or detecting social media influencer accounts that spread false information by combining with network causal inference could be implemented through deep learning. However, fake news detection is classified as a difficult problem to solve among many natural language processing fields. Due to the variety of forms and expressions of fake news, the difficulty of feature extraction is high, and there are various limitations, such as that one feature may have different meanings depending on the category to which the news belongs. In this paper, emotional change patterns are presented as an additional identification criterion for detecting fake news. We propose a model with improved performance by applying a convolutional neural network to a fake news data set to perform analysis based on content characteristics and additionally analyze emotional change patterns. Sentimental polarity is calculated for the sentences constituting the news and the result value dependent on the sentence order can be obtained by applying long-term and short-term memory. This is defined as a pattern of emotional change and combined with the content characteristics of news to be used as an independent variable in the proposed model for fake news detection. We train the proposed model and comparison model by deep learning and conduct an experiment using a fake news data set to confirm that emotion change patterns can improve fake news detection performance.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.4
• /
• pp.189-198
• /
• 2023

As crimes frequently occur on the street, the spread of CCTV is increasing. However, due to the shortcomings of passively operated CCTV, the need for intelligent CCTV is attracting attention. Due to the heavy system of such intelligent CCTV, high-performance devices are required, which has a problem in that it is expensive to replace the general CCTV. To solve this problem, an intelligent CCTV system that recognizes low-quality images and operates even on devices with low performance is required. Therefore, this paper proposes a Saying CCTV system that can detect threats in real time by using the AWS cloud platform to lighten the system and convert images into text. Based on the data extracted using YOLO v4 and OpenPose, it is implemented to determine the risk object, threat behavior, and threat situation, and calculate the risk using machine learning. Through this, the system can be operated anytime and anywhere as long as the network is connected, and the system can be used even with devices with minimal performance for video shooting and image upload. Furthermore, it is possible to quickly prevent crime by automating meaningful statistics on crime by analyzing the video and using the data stored as text.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.3
• /
• pp.141-148
• /
• 2023

Accurate overlay metrology is essential to achieve high yields of semiconductor products. Overlay metrology performance is greatly affected by overlay target design and measurement method. Therefore, in order to improve the performance of the overlay target, measurement methods applicable to various targets are required. In this study, we propose a new algorithm that can measure image-based overlay. The proposed measurement algorithm can estimate the sub-pixel position by using a motion vector. The motion vector may estimate the position of the sub-pixel unit by applying a quadratic equation model through polynomial expansion using pixels in the selected region. The measurement method using the motion vector can calculate the stacking error in all directions at once, unlike the existing correlation coefficient-based measurement method that calculates the stacking error on the X-axis and the Y-axis, respectively. Therefore, more accurate overlay measurement is possible by reflecting the relationship between the X-axis and the Y-axis. However, since the amount of computation is increased compared to the existing correlation coefficient-based algorithm, more computation time may be required. The purpose of this study is not to present an algorithm improved over the existing method, but to suggest a direction for a new measurement method. Through the experimental results, it was confirmed that measurement results similar to those of the existing method could be obtained.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.3
• /
• pp.133-140
• /
• 2023

The performance of lithium ion batteries depends on the usage environment and the combination ratio of cathode materials. In order to develop a high-performance lithium-ion battery, it is necessary to manufacture the battery and measure its performance while varying the cathode material ratio. However, it takes a lot of time and money to directly develop batteries and measure their performance for all combinations of variables. Therefore, research to predict the performance of a battery using an artificial intelligence model has been actively conducted. However, since measurement experiments were conducted with the same battery in the existing published battery data, the cathode material combination ratio was fixed and was not included as a data attribute. In this paper, we define a training data model required to develop an artificial intelligence model that can predict battery performance according to the combination ratio of cathode materials. We analyzed the factors that can affect the performance of lithium-ion batteries and defined the mass of each cathode material and battery usage environment (cycle, current, temperature, time) as input data and the battery power and capacity as target data. In the battery data in different experimental environments, each battery data maintained a unique pattern, and the battery classification model showed that each battery was classified with an error of about 2%.