Vol. 49, No. 11, Nov. 2022
Digital Library


Integration of Task Offloading, CPU Voltage Scaling, and Memory Placement for Power-Saving in Mobile Real-Time Systems

Soomin Ki, Gyuri Byun, Kyungwoon Cho, Hyokyung Bahn

http://doi.org/10.5626/JOK.2022.49.11.919

In this paper, we study real-time task scheduling that aims to minimize power consumption of CPU, memory, and network devices in mobile systems. By defining extended task model and adopting three low-power techniques (i.e., task offloading, CPU voltage scaling, and low-power memory placement), we co-optimize these three techniques, thereby saving power consumption of real-time systems by 76.8% on average. Our scheduling has the ability of rescheduling real-time task set by reconfiguring offloading, DVFS, and memory placement considering variations of network conditions, thereby minimizing power consumption without missing deadlines of given real-time tasks.

Exploring Text Summarization for Fake News Detection

Jie Bian, Seungeon Lee, Karandeep Singh, Meeyoung Cha

http://doi.org/10.5626/JOK.2022.49.11.926

Fake news detection models need to gather and ingest massive information from heterogeneous sources rapidly for solid verification. This paper demonstrates the feasibility of applying text summarization, to uncover useful information or evidence for fake news detection. Two popular deep learning-based summarization techniques, extractive and abstractive, were used to generate condensed textual information from lengthy news content. Experiments on popular rumor debunking datasets show that two lines of summarized text can extract critical information, while improving the classification performance and substantially reducing inference time. Text summarization can also bring explainability by providing evidence from three levels: words, sentences, and documents.

A Survey of High-Level Programming Languages for Operating Swarm Robots

Woosuk Kang, EunJin Jeong, Taejoon Yoon, Seokhaeng Heo, Soonhoi Ha

http://doi.org/10.5626/JOK.2022.49.11.936

Robots used for unmanned tasks in various fields have been developing in the direction of operating multiple robots rather than using only a single robot. Since swarm robotics aims to outperform a single robot through cooperative control of multiple robots, it can efficiently perform tasks such as building local maps and searching for survivors. Although developing and operating such swarm robot systems requires a skilled developer, high-level programming languages and frameworks that can support non-experts to easily specify and operate swarm robots are being developed. In this paper, we selected and introduced 11 swarm robot specification languages tested by simulations or using real robots. We compared the difference between selected languages from the viewpoint of software and swarm robot operation. Finally, limitations of swarm robotics languages developed so far and future tasks are discussed.

Detecting Implicitly Abusive Language by Applying Out-of-Distribution Problem

Jisu Shin, Hoyun Song, Jong C. Park

http://doi.org/10.5626/JOK.2022.49.11.948

Implicitly abusive language detection is a difficult problem to solve due to diversity of expressions and absence of a clear definition. Previous studies have claimed that implicitly abusive language should be classified and defined in detail, accompanied by corresponding datasets. However, this is not only inefficient, but also hard to flexibly respond to language changes. Our work proposes an efficient and effective method that processes implicitly abusive language as Out-of-Distribution data for the first time. In our experiments, a model with the proposed method performed better than a general pre-trained model and lexicon-based models. We also performed sentiment analysis and a case study to analyze characteristics of implicitly abusive language in detail and differences between a general pre-trained model and our model.

Layered Abstraction Technique for Effective Formal Verification of Deep Neural Networks

Jueun Yeon, Seunghyun Chae, Kyungmin Bae

http://doi.org/10.5626/JOK.2022.49.11.958

Deep learning has performed well in many areas. However, deep learning is vulnerable to errors such as adversarial examples. Therefore, much research exists on ensuring the safety and robustness of deep neural networks. Since deep neural networks are large in scale and the activation functions are non-linear, linear approximation methods for such activation functions are proposed and widely used for verification. In this paper, we propose a new technique, called layered abstraction, for non-linear activation functions, such as ReLU and Tanh, and the verification algorithm based on that. We have implemented our method by extending the existing SMT-based methods. The experimental evaluation showed that our tool performs better than an existing tool.

Solving Korean Math Word Problems Using the Graph and Tree Structure

Kwang Ho Bae, Sang Yeop Yeo, Yu Chul Jung

http://doi.org/10.5626/JOK.2022.49.11.972

In previous studies, there have been various efforts to solve math word problems in the English sentence system. In many studies, improved performance was achieved by introducing structures such as trees and graphs, beyond the Sequence-to-Sequence approaches. However, in the study of solving math problems in Korean sentence systems, there are no model cases, using structures such as trees or graphs. Thus, in this paper, we examine the possibility of solving math problems in Korean sentence systems for models using the tree structure, graph structure, and Korean pre-training language models together. Our experimental results showed that accuracy improved by approximately 20%, compared to the model of the Seq2seq structure, by introducing the graph and tree structure. Additionally, the use of the Korean pre-training language model showed an accuracy improvement of 4.66%-5.96%.

Effective Fuzzing Technique with Command Line Option Mutation/Selection

Ahcheong Lee, Yunho Kim, Moonzoo Kim

http://doi.org/10.5626/JOK.2022.49.11.981

Command line option determines how Command Line Interface (CLI) program executes. Therefore, selecting and using proper command line options determines the performance and effectiveness of testing in terms of coverage and bug detection power. In this paper, we propose PAW (Program option-AWare fuzzer) which automatically selects useful command line options to improve the effectiveness of fuzz testing. In the first 10% of fuzzing time, PAW generates diverse command line options by mutating existing options, and it decides which options are useful ones using branch coverage. For the remaining 90% of the time, PAW focuses on executing the selected, useful options for effective testing. We implemented PAW on top of AFL++ which is the state-of-the-art fuzzer, and we evaluated PAW on ten real-world C programs by comparing coverage with AFL++ and Angora. PAW showed higher branch coverage results and crash detection power on the ten C programs.

Recommendation Technique for Bug Fixers by Fine-tuning Language Models

Dae-Sung Wang, Hoon Seong, Chan-Gun Lee

http://doi.org/10.5626/JOK.2022.49.11.987

The scale and complexity of software continue to increase; hence they contribute to the occurrence of diverse bugs. Therefore, the necessity of systematic bug management has been raised. A few studies have proposed automating the assignment of bug fixers using word-based deep learning models. However, their accuracy is not satisfactory due to context of the word is ignored, and there is an excessive number of classes. In this paper, the accuracy was improved by about 27%p over the top-10 accuracies by using a fine-tuned pre-trained language model based on BERT, RoBERTa, DeBERTa, and CodeBERT. Experiments confirmed that the accuracy was about 70%. Through this, we showed that the fine-tuned pretrained language model could be effectively applied to automated bug-fixer assignments.

Explainable Graph Neural Network for Medical Science Research

Yewon Shin, Kisung Moon, Youngsuk Jung, Sunyoung Kwon

http://doi.org/10.5626/JOK.2022.49.11.999

Explainable AI (XAI) is a technology that provides explainability for our end-users to comprehend prediction results of ML algorithms. In particular, the reliability of the decision-making process of an AI algorithm through XAI technology is the most critical in the medical field in terms of real applications. However, complex interaction-based medical data restrict the application of existing XAI technologies developed mostly for image or text data. Graph Neural Network (GNN)-based XAI research has been highlighted in recent years because GNN is technically specialized to capture complex relationships in data. In this paper, we proposed a taxonomy according to the application method and algorithm of GNN-based XAI technology with current XAI research trends and its use-cases in four detailed areas of the medical field. We also expounded on the technical limitations and future works of XAI research specialized in the biomedical area.

Building a Parallel Corpus and Training Translation Models Between Luganda and English

Richard Kimera, Daniela N. Rim, Heeyoul Choi

http://doi.org/10.5626/JOK.2022.49.11.1009

Recently, neural machine translation (NMT) which has achieved great successes needs large datasets, so NMT is more premised on high-resource languages. This continuously underpins the low resource languages such as Luganda due to the lack of high-quality parallel corpora, so even ‘Google translate’ does not serve Luganda at the time of this writing. In this paper, we build a parallel corpus with 41,070 pairwise sentences for Luganda and English which is based on three different open-sourced corpora. Then, we train NMT models with hyper-parameter search on the dataset. Experiments gave us a BLEU score of 21.28 from Luganda to English and 17.47 from English to Luganda. Some translation examples show high quality of the translation. We believe that our model is the first Luganda-English NMT model. The bilingual dataset we built will be available to the public.

Korean-English Neural Machine Translation Using Korean Alphabet Characteristics and Honorific Expressions

Jeonghui Kim, Jaemu Heo, Joowhan Kim, Heeyoul Choi

http://doi.org/10.5626/JOK.2022.49.11.1017

Recently, deep learning has improved the performance of machine translation, but in most cases, it does not reflect the characteristics of the languages. In particular, Korean has unique linguistic word and expression features, which might cause mistranslation. For example, in Google Translate from Korean to English, mistranslations occur when a noun in Korean ends with the postposition (josa) in the form of a single consonant. Also, in the English-Korean translations, the honorifics and casual expressions are mixed in the translated results. This is because the alphabetic characteristics and honorifics of the Korean language are not reflected. In this paper, to address these problems, we propose to train a model with sub-words composed of units of letters (jamo) and unifying honorific and casual expressions in the corpus. The experimental results confirmed that the proposed method resolved the problems mentioned above, and had a similar or slightly higher BLEU score compared to the existing method and the corpus.

OANet: Ortho-Attention Net Based on Attention Mechanism for Database Performance Prediction

Chanho Yeom, Jieun Lee, Sanghyun Park

http://doi.org/10.5626/JOK.2022.49.11.1026

Various parameters in a database can be modified, which are called knobs. Since the performance of the database varies according to the settings of the knobs, it is important to tune the knobs of the database. And when tuning, a model that can reliably and quickly predict database performance according to the knob setting is needed. However, even when the knob setting is the same, the results may be different if the workload performing the benchmark is different. Therefore, in this paper, we propose an OANet using the attention mechanism so that the relationship between the knob and the workload can also be considered. Through experiments, the performance prediction results of the database were compared to various machine learning techniques, and the superiority of the model was confirmed by showing the highest score.

Non-Fungible Token-based Virtual Avatar for User Identification in the Metaverse: Methodology and User Study for Virtual Reality Social Interaction

Jihwan Kim, Kwanguk (Kenny) Kim

http://doi.org/10.5626/JOK.2022.49.11.1032

Anonymity in the metaverse is a property that allows users to act freely in the virtual world, but it also causes several problems, including reliability with socially interactive partners. These problems can be partially addressed by traditional methodologies, but also require a more advanced method. To address this issue, we proposed a non-fungible token (NFT)-based avatar with the user"s appearance and personal information. A user study of 31 participants was also conducted to compare the proposed NFT-based avatar and traditional avatars in a socially interactive context. We found that higher interpersonal trust, higher co-presences, and positive emotional responses were formed when a user interacted with NFT-based avatars. This study has contributions in that we proposed an NFT-based virtual avatar methodology with user attributes and verified its usability in a socially interactive context.


Search




Journal of KIISE

  • ISSN : 2383-630X(Print)
  • ISSN : 2383-6296(Electronic)
  • KCI Accredited Journal

Editorial Office

  • Tel. +82-2-588-9240
  • Fax. +82-2-521-1352
  • E-mail. chwoo@kiise.or.kr