• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.55-60
• /
• 2021

최근 한국어에 대한 자연어 처리 연구는 딥러닝 기반의 자연어 이해 모델을 중심으로 각 모델의 성능에 대한 비교 분석과 평가가 활발하게 이루어지고 있다. 그러나 한국어 생성 모델에 대해서도 자연어 이해 영역의 하위 과제(e.g. 감정 분류, 문장 유사도 측정 등)에 대한 수행 능력만을 정량적으로 평가하여, 생성 모델의 한국어 문장 구성 능력이나 상식 추론 과정을 충분히 평가하지 못하고 있다. 또한 대부분의 생성 모델은 여전히 간단하고 일반적인 상식에 부합하는 자연스러운 문장을 생성하는 것에도 큰 어려움을 겪고 있기에 이를 해결하기 위한 개선 연구가 필요한 상황이다. 따라서 본 논문은 이러한 문제를 해결하기 위해 한국어 생성 모델이 일반 상식 추론 능력을 바탕으로 문장을 생성하도록 KommonGen 데이터셋을 제안한다. 그리고 KommonGen을 통해 한국어 생성 모델의 성능을 정량적으로 비교 분석할 수 있도록 평가 기준을 구성하고, 한국어 기반 자연어 생성 모델의 개선 방향을 제시하고자 한다.

• Journal of Convergence for Information Technology
• /
• v.11 no.6
• /
• pp.49-54
• /
• 2021

With the recent development of smart healthcare technology, interest in daily diseases is increasing. However, healthcare data has an imbalance between positive and negative data. This is caused by the difficulty of collecting data because there are relatively many people who are not patients compared to patients with certain diseases. Data imbalances need to be adjusted because they affect performance in ongoing learning during disease prediction and analysis. Therefore, in this paper, We replace missing values through multiple imputation in detection models to determine whether they are prevalent or not, and resolve data imbalances through over-sampling. Based on AutoML using preprocessed data, We generate several models and select top 3 models to generate ensemble models.

• Journal of Intelligence and Information Systems
• /
• v.28 no.1
• /
• pp.217-242
• /
• 2022

With the launch of Artificial Intelligence(AI)-based intelligent products on the market, innovative changes are taking place not only in business but also in consumers' daily lives. Intelligent products have the potential to realize technology differentiation and increase market competitiveness through advanced functions of artificial intelligence. However, there is no new product development methodology that can sufficiently reflect the characteristics of artificial intelligence for the purpose of developing intelligent products with high market acceptance. This study proposes a KANO-QFD integrated model as a methodology for intelligent product development. As a specific example of the empirical analysis, the types of consumer requirements for hair loss prediction and treatment device were classified, and the relative importance and priority of engineering characteristics were derived to suggest the direction of intelligent medical product development. As a result of a survey of 130 consumers, accurate prediction of future hair loss progress, future hair loss and improved future after treatment realized and viewed on a smartphone, sophisticated design, and treatment using laser and LED combined light energy were realized as attractive quality factors among the KANO categories. As a result of the analysis based on House of Quality of QFD, learning data for hair loss diagnosis and prediction, micro camera resolution for scalp scan, hair loss type classification model, customized personal account management, and hair loss progress diagnosis model were derived. This study is significant in that it presented directions for the development of artificial intelligence-based intelligent medical product that were not previously preceded.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.2
• /
• pp.21-30
• /
• 2024

This research aims to design a system capable of generating real web pages based on deep learning and big data, in three stages. First, a classification system was established based on the industry type and functionality of e-commerce websites. Second, the types of components of web pages were systematically categorized. Third, the entire web page auto-generation system, applicable for deep learning, was designed. By re-engineering the deep learning model, which was trained with actual industrial data, to analyze and automatically generate existing websites, a directly usable solution for the field was proposed. This research is expected to contribute technically and policy-wise to the field of generative AI-based complete website creation and industrial sectors.

• The Transactions of the Korea Information Processing Society
• /
• v.13 no.5
• /
• pp.209-216
• /
• 2024

Real estate market prices are determined by various factors, including macroeconomic variables, as well as the influence of a variety of unstructured text data such as news articles and social media. News articles are a crucial factor in predicting real estate transaction prices as they reflect the economic sentiment of the public. This study utilizes sentiment analysis on news articles to generate a News Sentiment Index score, which is then seamlessly integrated into a real estate price prediction model. To calculate the sentiment index, the content of the articles is first summarized. Then, using AI, the summaries are categorized into positive, negative, and neutral sentiments, and a total score is calculated. This score is then applied to the real estate price prediction model. The models used for real estate price prediction include the Multi-head attention LSTM model and the Vector Auto Regression model. The LSTM prediction model, without applying the News Sentiment Index (NSI), showed Root Mean Square Error (RMSE) values of 0.60, 0.872, and 1.117 for the 1-month, 2-month, and 3-month forecasts, respectively. With the NSI applied, the RMSE values were reduced to 0.40, 0.724, and 1.03 for the same forecast periods. Similarly, the VAR prediction model without the NSI showed RMSE values of 1.6484, 0.6254, and 0.9220 for the 1-month, 2-month, and 3-month forecasts, respectively, while applying the NSI led to RMSE values of 1.1315, 0.3413, and 1.6227 for these periods. These results demonstrate the effectiveness of the proposed model in predicting apartment transaction price index and its ability to forecast real estate market price fluctuations that reflect socio-economic trends.

• Journal of Internet Computing and Services
• /
• v.24 no.5
• /
• pp.17-27
• /
• 2023

Today, as AI (Artificial Intelligence) technology develops and its practicality increases, it is widely used in various application fields in real life. At this time, the AI model is basically learned based on various statistical properties of the learning data and then distributed to the system, but unexpected changes in the data in a rapidly changing data situation cause a decrease in the model's performance. In particular, as it becomes important to find drift signals of deployed models in order to respond to new and unknown attacks that are constantly created in the security field, the need for lifecycle management of the entire model is gradually emerging. In general, it can be detected through performance changes in the model's accuracy and error rate (loss), but there are limitations in the usage environment in that an actual label for the model prediction result is required, and the detection of the point where the actual drift occurs is uncertain. there is. This is because the model's error rate is greatly influenced by various external environmental factors, model selection and parameter settings, and new input data, so it is necessary to precisely determine when actual drift in the data occurs based only on the corresponding value. There are limits to this. Therefore, this paper proposes a method to detect when actual drift occurs through an Anomaly analysis technique based on XAI (eXplainable Artificial Intelligence). As a result of testing a classification model that detects DGA (Domain Generation Algorithm), anomaly scores were extracted through the SHAP(Shapley Additive exPlanations) Value of the data after distribution, and as a result, it was confirmed that efficient drift point detection was possible.

• Journal of Intelligence and Information Systems
• /
• v.29 no.4
• /
• pp.15-30
• /
• 2023

Recently, various services using artificial intelligence(AI) are emerging in the media field as well However, most of the video editing, which involves finding an editing point and attaching the video, is carried out in a passive manner, requiring a lot of time and human resources. Therefore, this study proposes a methodology that can detect the edit points of video according to whether person in video are spoken by using Video Swin Transformer. First, facial keypoints are detected through face alignment. To this end, the proposed structure first detects facial keypoints through face alignment. Through this process, the temporal and spatial changes of the face are reflected from the input video data. And, through the Video Swin Transformer-based model proposed in this study, the behavior of the person in the video is classified. Specifically, after combining the feature map generated through Video Swin Transformer from video data and the facial keypoints detected through Face Alignment, utterance is classified through convolution layers. In conclusion, the performance of the image editing point detection model using facial keypoints proposed in this paper improved from 87.46% to 89.17% compared to the model without facial keypoints.

• Annual Conference of KIPS
• /
• 2023.11a
• /
• pp.611-614
• /
• 2023

갈비뼈 병변 진단 과정은 방사선 전문의가 CT 스캐너를 통해 생성된 2 차원 CT 이미지들을 해석하며 진행된다. 병변의 위치를 파악하고 정확한 진단을 내리기 위해 수백장의 2차원 CT 이미지들이 세밀하게 검토되며 갈비뼈를 분류한다. 본 연구는 이런 노동 집약적 작업의 문제점을 개선시키기 위해 Biaxial Rib Segmentation(BARS)을 제안한다. BARS 는 흉부 CT 볼륨의 관상면과 수평면으로 구성된 2 차원 이미지들을 U-Net 모델에 학습한다. 모델이 산출한 세그멘테이션 마스크들의 조합은 서로 다른 평면의 공간 정보를 보완하며 3 차원 갈비뼈 볼륨을 재건한다. BARS 의 성능은 DSC, Recall, Precision 지표를 사용해 평가하며, DSC 90.29%, Recall 89.74%, Precision 90.72%를 보인다. 향후에는 이를 기반으로 순차적 갈비뼈 레이블링 연구를 진행할 계획이다.

• Annual Conference of KIPS
• /
• 2023.11a
• /
• pp.597-598
• /
• 2023

X-ray는 촬영 방식의 한계로 진단하기 어려운, 품질 낮은 영상을 다수 발생시킨다. 이러한 저품질 영상은 임상 현장에서의 진단이 어려울 뿐만 아니라, 진단 보조 도구를 개발함에 모델의 성능과 신뢰도를 떨어뜨리는 주요 요소가 된다. 특히 투과도가 낮은 영상은 학습 성능에 악영향을 미친다는 것이 입증된 바 있다. 따라서 본 연구는 투과도가 낮은 영상을 진단에 부적합한 영상으로 정의하여, 이를 분류하는 기법을 제안한다. 제안하는 기법은 민감도 94.9%. 특이도 96.0%의 높은 성능을 보였다.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.21 no.6
• /
• pp.730-736
• /
• 2011

In this paper, we designed the models for pattern classification which can reflect the latest trend in time series. It has been shown that fusion models based on statistical and AI methods are superior to traditional ones for the pattern classification model supporting decision making. Especially, the hit rates of pattern classification models combined with fuzzy theory are relatively increased. The statistical SVM models combined with fuzzy membership function, or the models combining neural network and FCM has shown good performance. BPN, PNN, FNN, FCM, SVM, FSVM, Decision Tree, Time Series Analysis, and Regression Analysis were used for pattern classification models in the experiments of this paper. The economical indices DB with time series properties of the financial market(Korea, KOSPI200 DB) and the electrocardiogram DB of arrhythmia patients in hospital emergencies(USA, MIT-BIH DB) were used for data base.