• Knowledge Management Research
• /
• v.24 no.4
• /
• pp.153-176
• /
• 2023

Generative AI, especially conversational AI like ChatGPT, has recently gained traction as a technological alternative for automating customer service. However, there is still a lack of research on whether current generative AI technologies can effectively replace traditional human managers in customer service automation, and whether they are advantageous in some situations and disadvantageous in others, depending on the conditions and environment. To answer the question, "Can generative AI replace human managers in customer service activities?", this study conducted experiments and surveys on customer online reviews of a food delivery platform. We applied the perspective of the elaboration likelihood model to generate hypotheses about whether there is a difference between positive and negative online reviews, and analyzed whether the hypotheses were supported. The analysis results indicate that for positive reviews, generative AI can effectively replace human managers. However, for negative reviews, complete replacement is challenging, and human managerial intervention is considered more desirable. The results of this study can provide valuable practical insights for organizations looking to automate customer service using generative AI.

• Journal of Sensor Science and Technology
• /
• v.33 no.2
• /
• pp.112-116
• /
• 2024

Gas-sensor technology for volatile organic compounds (VOC) biomarker detection offers significant advantages for noninvasive diagnostics, including rapid response time and low operational costs, exhibiting promising potential for disease diagnosis. Colorimetric gas sensors, which enable intuitive analysis of gas concentrations through changes in color, present additional benefits for the development of personal diagnostic kits. However, the traditional method of visually monitoring these sensors can limit quantitative analysis and consistency in detection threshold evaluation, potentially affecting diagnostic accuracy. To address this, we developed a machine vision platform based on metal-organic framework (MOF) for colorimetric gas sensor arrays, designed to accurately detect disease-related VOC biomarkers. This platform integrates a CMOS camera module, gas chamber, and colorimetric MOF sensor jig to quantitatively assess color changes. A specialized machine vision algorithm accurately identifies the color-change Region of Interest (ROI) from the captured images and monitors the color trends. Performance evaluation was conducted through experiments using a platform with four types of low-concentration standard gases. A limit-of-detection (LoD) at 100 ppb level was observed. This approach significantly enhances the potential for non-invasive and accurate disease diagnosis by detecting low-concentration VOC biomarkers and offers a novel diagnostic tool.

• Journal of Internet of Things and Convergence
• /
• v.10 no.2
• /
• pp.125-130
• /
• 2024

With the rapid growth of mobility technology, the industrial sector is demanding storage devices that can reliably process data from various equipment and sensors in vehicles. NAND flash memory is being utilized as a storage device in mobility environments because it has the advantages of low power and fast data processing speed as well as strong external shock resistance. However, flash memory is characterized by data corruption due to long-term exposure to high temperatures. Therefore, a dedicated system for temperature management is required in mobility environments where high temperature exposure due to weather or external heat sources such as solar radiation is frequent. This paper designs a dedicated temperature management system for managing storage device temperature in a mobility environment. The designed temperature management system is a hybrid of traditional air cooling and water cooling technologies. The cooling method is designed to operate adaptively according to the temperature of the storage device, and it is designed not to operate when the temperature step is low to improve energy efficiency. Finally, experiments were conducted to analyze the temperature difference between each cooling method and different heat dissipation materials, proving that the temperature management policy is effective in maintaining performance.

• Journal of Korea Society of Industrial Information Systems
• /
• v.29 no.2
• /
• pp.15-25
• /
• 2024

For effective analysis of animal ecosystems, technology that can automatically identify the current status of animal habitats is crucial. Specifically, animal sound classification, which identifies species based on their sounds, is gaining great attention where video-based discrimination is impractical. Traditional studies have relied on a single deep learning model to classify animal sounds. However, sounds collected in outdoor settings often include substantial background noise, complicating the task for a single model. In addition, data imbalance among species may lead to biased model training. To address these challenges, in this paper, we propose an animal sound classification scheme that combines predictions from multiple models using Focal Loss, which adjusts penalties based on class data volume. Experiments on public datasets have demonstrated that our scheme can improve recall by up to 22.6% compared to an average of single models.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.2
• /
• pp.348-369
• /
• 2024

With the advancement of Industry 4.0 and Industrial Internet of Things (IIoT), manufacturing increasingly seeks automation and intelligence. Temperature and vibration monitoring are essential for machinery health. Traditional abnormal state detection methodologies often overlook the intricate frequency characteristics inherent in vibration time series and are susceptible to erroneously reconstructing temperature abnormalities due to the highly similar waveforms. To address these limitations, we introduce synergistic, end-to-end, unsupervised Frequency-Time Domain Memory-Enhanced Autoencoders (FTD-MAE) capable of identifying abnormalities in both temperature and vibration datasets. This model is adept at accommodating time series with variable frequency complexities and mitigates the risk of overgeneralization. Initially, the frequency domain encoder processes the spectrogram generated through Short-Time Fourier Transform (STFT), while the time domain encoder interprets the raw time series. This results in two disparate sets of latent representations. Subsequently, these are subjected to a memory mechanism and a limiting function, which numerically constrain each memory term. These processed terms are then amalgamated to create two unified, novel representations that the decoder leverages to produce reconstructed samples. Furthermore, the model employs Spectral Entropy to dynamically assess the frequency complexity of the time series, which, in turn, calibrates the weightage attributed to the loss functions of the individual branches, thereby generating definitive abnormal scores. Through extensive experiments, FTD-MAE achieved an average ACC and F1 of 0.9826 and 0.9808 on the CMHS and CWRU datasets, respectively. Compared to the best representative model, the ACC increased by 0.2114 and the F1 by 0.1876.

• The Korean Journal of Pain
• /
• v.37 no.2
• /
• pp.141-150
• /
• 2024

Background: Stingless bee propolis is a popular traditional folk medicine and has been employed since ancient times. This study aimed to evaluate the antinociceptive activities of the chemical constituents of aqueous propolis extract (APE) collected by Trigona thoracica in a nociceptive model in mice. Methods: The identification of chemical constituents of APE was performed using high-performance liquid chromatography (HPLC). Ninety-six male Swiss mice were administered APE (400 mg/kg, 1,000 mg/kg, and 2,000 mg/kg) before developing nociceptive pain models. Then, the antinociceptive properties of each APE dose were evaluated in acetic acid-induced abdominal constriction, hot plate test, and formalin-induced paw licking test. Administration of normal saline, acetylsalicylic acid (ASA, 100 mg/kg, orally), and morphine (5 mg/kg, intraperitoneally) were used for the experiments. Results: HPLC revealed that the APE from Trigona thoracica contained p-coumaric acid (R² = 0.999) and caffeic acid (R² = 0.998). Although all APE dosages showed inhibition of acetic acid-induced abdominal constriction, only 2,000 mg/kg was comparable to the result of ASA (68.7% vs. 73.3%, respectively). In the hot plate test, only 2,000 mg/kg of APE increased the latency time significantly compared to the control. In the formalin test, the durations of paw licking were significantly reduced at early and late phases in all APE groups with a decrease from 45.1% to 53.3%. Conclusions: APE from Trigona thoracica, containing p-coumaric acid and caffeic acid, exhibited antinociceptive effects, which supports its potential use in targeting the prevention or reversal of central and peripheral sensitization that may produce clinical pain conditions.

• Convergence Security Journal
• /
• v.23 no.4
• /
• pp.61-70
• /
• 2023

Clustering is an unsupervised learning method that involves grouping data based on features such as distance metrics, using data without known labels or ground truth values. This method has the advantage of being applicable to various types of data, including images, text, and audio, without the need for labeling. Traditional clustering techniques involve applying dimensionality reduction methods or extracting specific features to perform clustering. However, with the advancement of deep learning models, research on deep clustering techniques using techniques such as autoencoders and generative adversarial networks, which represent input data as latent vectors, has emerged. In this study, we propose a deep clustering technique based on deep learning. In this approach, we use an autoencoder to transform the input data into latent vectors, and then construct a vector space according to the cluster structure and perform k-means clustering. We conducted experiments using the MNIST and Fashion-MNIST datasets in the PyTorch machine learning library as the experimental environment. The model used is a convolutional neural network-based autoencoder model. The experimental results show an accuracy of 89.42% for MNIST and 56.64% for Fashion-MNIST when k is set to 10.

• The Journal of Korean Medicine Ophthalmology and Otolaryngology and Dermatology
• /
• v.37 no.1
• /
• pp.17-28
• /
• 2024

Objectives : This study was conducted to reveal the scientific mechanism of the anti-skin aging activity of Kyung-Ok-Ko(KOK), which is highly useful as a Korean traditional medicine and functional food. Methods : The skin wrinkle and aging inhibitory activity of KOK was confirmed through in vitro experiments of human dermal fibroblast neonatal cell(HDFn) and in vivo of C. elegans, and hairless mouse(SKH-1). Results : The amount of the C-terminus of the collagen precursor in the HDFn cell culture medium treated with KOK using an enzymes-linked immunoassay kit. The group treated with KOK 200㎍/㎖ was a 28.3% increase of collagen precursor compared to the control group. KOK showed inhibitory activity of MMP-1 compared to the control group at a concentration of 200㎍/㎖. In addition, KOK 200㎍/㎖ showed significant inhibitory activity of thermal stress and an oxidative stress compared to the control group in C. elegans. Furthermore, KOK showed a concentration-dependent(100mg/kg and 500mg/kg) anti-wrinkle formation effect in UV-irradiated hairless mouse(SKH-1). Additionally, when KOK was administered to UV-irradiated hairless mice, an increase in procollagen -1 and -3 genes expression was observed, and mmp-1 and mmp-9 genes, which increase collagen decomposition, decreased with the administration of KOK. Conclusions : The skin aging inhibition mechanism of Kyung-Ok-Ko(KOK) is presumed to be achieved through suppressing thermal stress and oxidative stress, suppressing mmp-1 and mmp-9 genes, and increasing procollagen-1 and procollagen-3.

• The Journal of Korean Medicine Ophthalmology and Otolaryngology and Dermatology
• /
• v.37 no.1
• /
• pp.57-68
• /
• 2024

Objectives : The purpose of this study is to analyze the current trends of various herbal medicine research on allergic disease with dysbiosis. Methods : Electronic searches were performed using Pubmed, Research Information Sharing Service(RISS), Korean studies Information Service System(KISS), Oriental medicine Advanced Searching Integrated System(OASIS). Results : We analyzed ten studies on the effect of herbal medicine on allergic disease with dysbiosis. Eight studies were animal experimental studies, and two were randomized clinical trial(RCT) study and one-group pretest-posttest research, respectively. Among the studies, three studies were on atopic dermatitis, two on allergic rhinitis, and five on asthma. All different herbal medicines were used in the studies. Changes in gut microbiota composition were observed in nine studies except for 1 RCT study. In eight animal experimental studies, there was significant reduction in allergy-related inflammatory markers. Six studies evaluated the change of metabolites related to gut microbiota and three of them showed significant increase in short-chain fatty acids(SCFA). Conclusion : This study provides current trends of studies on herbal medicine research on allergic disease with dysbiosis. Most research is conducted using animal experiments, and this is a relatively recent trend. These studies offer basic knowledge on the correlation between herbal medicine, gut microbiota, and anti-inflammatory effects in allergic disease.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.4
• /
• pp.9-21
• /
• 2024

Deep learning has been developing rapidly in recent years, with many researchers working to utilize large language models in various domains. However, there are practical difficulties that developing and utilizing language models require massive data and high-performance computing resources. Therefore, in-context learning, which utilizes prompts to learn efficiently, has been introduced, but there needs to be clear criteria for effective prompts for learning. In this study, we propose a methodology for enhancing prompt-based learning performance by improving the PET technique, which is one of the contextual learning methods, to select PVPs that are similar to the context of existing data. To evaluate the performance of the proposed methodology, we conducted experiments with 30,100 restaurant review datasets collected from Yelp, an online business review platform. We found that the proposed methodology outperforms traditional PET in all aspects of accuracy, stability, and learning efficiency.