• 의학교육논단
• /
• 제19권3호
• /
• pp.158-168
• /
• 2017

There is a more urgent call for educational methods of machine learning in medical education, and therefore, new approaches of teaching and researching machine learning in medicine are needed. This paper presents a case using machine learning through text analysis. Topic modeling of news articles with the keyword 'asbestos' were examined. Two hypotheses were tested using this method, and the process of machine learning of texts is illustrated through this example. Using an automated text analysis method, all the news articles published from January 1, 1990 to November 15, 2016 in South Korea which included 'asbestos' in the title and the body were collected by web scraping. Differences in topics were analyzed by structured topic modelling (STM) and compared by press companies and periods. More articles were found in liberal media outlets. Differences were found in the number and types of topics in the articles according to the partisanship and period. STM showed that the conservative press views asbestos as a personal problem, while the progressive press views asbestos as a social problem. A divergence in the perspective for emphasizing the issues of asbestos between the conservative press and progressive press was also found. Social perspective influences the main topics of news stories. Thus, the patients' uneasiness and pain are not presented by both sources of media. In addition, topics differ between news media sources based on partisanship, and therefore cause divergence in readers' framing. The method of text analysis and its strengths and weaknesses are explained, and an application for the teaching and researching of machine learning in medical education using the methodology of text analysis is considered. An educational method of machine learning in medical education is urgent for future generations.

• 한국비블리아학회지
• /
• 제32권3호
• /
• pp.247-264
• /
• 2021

자살은 전 세계 사망 원인 중 4위이며 사회, 경제적 손실이 큰 난제이다. 본 연구는 자살 예방을 위하여 소셜미디어에 나타난 자살 관련 말뭉치를 구축하고 이를 통해 자살 경향 문헌을 분류할 수 있는 딥러닝 자동분류 모델을 만들고자 하였다. 또한, 자살 요인을 분석하기 위해 주제를 자동으로 추출하는 분석 기법인 토픽모델링을 활용하여 자살 관련 말뭉치를 세부 주제로 분류하고자 하였다. 이를 위해 소셜미디어 중 하나인 네이버 지식iN에 나타난 자살 관련 문헌 2,011개를 수집한 후 자살예방교육 매뉴얼을 기준으로 자살 경향 문헌 및 비경향 문헌 여부를 주석 처리하였으며, 이 데이터를 딥러닝 모델(LSTM, BERT, ELECTRA)로 학습시켜 자동분류 모델을 만들었다. 또한, 토픽모델링 기법의 하나인 LDA 기법으로 주제별 문헌을 분류하여 자살 요인을 발견하였고 이를 심층적으로 분석하기 위해 주제별로 동시출현 단어 분석 및 네트워크 시각화를 진행하였다.

• International Journal of Computer Science & Network Security
• /
• 제24권2호
• /
• pp.79-88
• /
• 2024

The use of the social media has become part of our daily life activities. The social web channels provide the content generation facility to its users who can share their views, opinions and experiences towards certain topics. The researchers are using the social media content for various research areas. Sentiment analysis, one of the most active research areas in last decade, is the process to extract reviews, opinions and sentiments of people. Sentiment analysis is applied in diverse sub-areas such as subjectivity analysis, polarity detection, and emotion detection. Stance classification has emerged as a new and interesting research area as it aims to determine whether the content writer is in favor, against or neutral towards the target topic or issue. Stance classification is significant as it has many research applications like rumor stance classifications, stance classification towards public forums, claim stance classification, neural attention stance classification, online debate stance classification, dialogic properties stance classification etc. This research study explores different feature sets such as lexical, sentiment-specific, dialog-based which have been extracted using the standard datasets in the relevant area. Supervised learning approaches of generative algorithms such as Naïve Bayes and discriminative machine learning algorithms such as Support Vector Machine, Naïve Bayes, Decision Tree and k-Nearest Neighbor have been applied and then ensemble-based algorithms like Random Forest and AdaBoost have been applied. The empirical based results have been evaluated using the standard performance measures of Accuracy, Precision, Recall, and F-measures.

• 로봇학회논문지
• /
• 제10권2호
• /
• pp.90-95
• /
• 2015

The main goal of e-learning systems is just-in-time knowledge acquisition. Rule-based e-learning systems, however, suffer from the mesa effect and the cold start problem, which both result in low user acceptance. E-learning systems suffer a further drawback in rendering the implementation of a natural interface in humanoids difficult. To address these concerns, even exceptional questions of the learner must be answerable. This paper aims to propose a method that can understand the learner's verbal cues and then intelligently explore additional domains of knowledge based on crowd data sources such as Wikipedia and social media, ultimately allowing for better answers in real-time. A prototype system was implemented using the NAO platform.

• 스마트미디어저널
• /
• 제11권11호
• /
• pp.25-31
• /
• 2022

자율주행 자동차에서 딥러닝 기반 영상처리는 매우 중요하다. 자동차를 비롯한 SoC(System on Chip) 환경에서 실시간으로 도로 영상을 처리하기 위해서는 영상처리 모델을 딥러닝 연산에 특화된 NPU(Neural Processing Unit) 상에서 실행해야 한다. 본 연구에서는 GPU 서버 환경에서 개발된 7종의 오픈소스 딥러닝 영상처리 모델들을 TIDL (Texas Instrument Deep Learning) NPU 환경에 이식하였다. 성능 평가와 시각화를 통해 본 연구에서 이식한 모델들이 SoC 가상환경에서 정상 작동함을 확인하였다. 본 논문은 NPU 환경의 제약으로 인해 이식 과정에 발생한 문제들과 그 해결 방법을 소개함으로써 딥러닝 모델을 SoC 환경에 이식하려는 개발자 및 연구자가 참고할 만한 사례를 제시한다.

• 스마트미디어저널
• /
• 제13권3호
• /
• pp.27-35
• /
• 2024

본 연구는 GAN(Generative Adversarial Network)을 기반으로 한 Semi Supervised Learning을 활용하여 이미지 생성과 ResNet50을 이용한 이미지 분류를 결합하는 방법에 대해 다루고 있다. 이를 통해 새로운 접근법을 제시하여 이미지 생성과 분류를 통합함으로써 더 정확하고 다양한 결과를 얻을 수 있도록 하였다. 생성자와 판별자를 학습시켜 생성된 이미지와 실제 이미지를 구별하고, ResNet50을 활용하여 이미지 분류를 수행한다. 실험 결과에서는 생성된 이미지의 품질이 epoch에 따라 변화함을 확인할 수 있었으며, 이를 통해 산업재해 예측 정확성을 향상하고자 한다. 또한, GAN과 ResNet50의 결합을 통해 이미지 생성의 품질을 향상시키고 이미지 분류의 정확도를 높이는 효율적인 방법을 제시하고자 한다.

• 스마트미디어저널
• /
• 제13권8호
• /
• pp.16-25
• /
• 2024

Recently, there has been a growing emphasis on identifying both known and unknown diseases in plant disease recognition. In this task, a model trained only on images of known classes is required to classify an input image into either one of the known classes or into an unknown class. Consequently, the capability to recognize unknown diseases is critical for model deployment. To enhance this capability, we are considering three factors. Firstly, we propose a new logits-based scoring function for unknown scores. Secondly, initial experiments indicate that a compact feature space is crucial for the effectiveness of logits-based methods, leading us to employ the AM-Softmax loss instead of Cross-entropy loss during training. Thirdly, drawing inspiration from the efficacy of transfer learning, we utilize a large plant-relevant dataset, PlantCLEF2022, for pre-training a model. The experimental results suggest that our method outperforms current algorithms. Specifically, our method achieved a performance of 97.90 CSA, 91.77 AUROC, and 90.63 OSCR with the ResNet50 model and a performance of 98.28 CSA, 92.05 AUROC, and 91.12 OSCR with the ConvNext base model. We believe that our study will contribute to the community.

• 스마트미디어저널
• /
• 제6권3호
• /
• pp.41-48
• /
• 2017

Event detection using social media has been widespread since social network services have been an active communication channel for connecting with others, diffusing news message. Especially, the real-time characteristic of social media has created the opportunity for supporting for real-time applications/systems. Social network such as Twitter is the potential data source to explore useful information by mining messages posted by the user community. This paper proposed a novel system for temporal event detection by analyzing social data. As a result, this information can be used by first responders, decision makers, or news agents to gain insight of the situation. The proposed approach takes advantages of deep learning methods that play core techniques on the main tasks including informative data identifying from a noisy environment and temporal event detection. The former is the responsibility of Convolutional Neural Network model trained from labeled Twitter data. The latter is for event detection supported by Recurrent Neural Network module. We demonstrated our approach and experimental results on the case study of earthquake situations. Our system is more adaptive than other systems used traditional methods since deep learning enables to extract the features of data without spending lots of time constructing feature by hand. This benefit makes our approach adaptive to extend to a new context of practice. Moreover, the proposed system promised to respond to acceptable delay within several minutes that will helpful mean for supporting news channel agents or belief plan in case of disaster events.

• 만화애니메이션 연구
• /
• 통권47호
• /
• pp.295-333
• /
• 2017

본 연구는 교과용도서 내 영상물 수록이 활발해짐에 따라 영상물 선정의 기준과 관련된 정책을 마련하기 위하여 실행된 기초연구이다. 이를 위해 먼저 영상물이 어린이 및 청소년의 발달단계에 따른 정서적 영향에 대한 연구를 살펴보고 영상물과 학생의 정서 및 건강, 교육적 효과성 간의 관련성을 밝히고자 하였다. 이후 국내외 영상물 관련 심의 및 등급분류 기준을 폭넓게 검토함으로써 국가 수준의 정책 차원에서 영상물 등급제를 제도화 할 필요가 있음을 주장하였다. 위의 사항을 바탕으로 연구자들은 일곱 가지 제언을 하였다. 첫째, 교과용도서 편찬상의 유의점 및 편수자료 등에 영상물 선정 기준을 명시할 필요가 있다. 둘째, 교과용도서에 수록하기 위한 영상물의 정치적 중립성과 인권 측면을 검토하는 데 도움이 되는 지침이 필요하다. 셋째, 국내외 영상물 등급 제도의 범주 항목 및 연령별 준거를 참고하여 교과용도서 내 영상물 선정 지침을 상세화해야 한다. 넷째, 명백한 교육적 목적이 있을 경우에 한하여 영상물 등급 제도를 유연하게 적용할 수 있도록 한다. 다섯째, 교과용도서의 영상물 수록 지침 설정을 위한 제도적 지원이 필요하다. 여섯째, 교과용도서 개발 전 과정에 영상물 전문가 집단이 참여해야 한다. 일곱째, 교실 수업에서 교육용 영상물을 활용하여 자기주도적 학습을 할 수 있도록 교사 교육 프로그램을 병행해야 한다.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2022년도 추계학술대회
• /
• pp.112-115
• /
• 2022

The process of deep learning usually needs to deal with massive data which has greatly limited the development of deep learning technologies today. Convolutional Neural Network (CNN) structure is often used to solve image classification problems. However, a large number of images may be required in order to train an image in CNN, which is a heavy burden for existing computer systems to handle. If the image data can be compressed under the premise that the computer hardware system remains unchanged, it is possible to train more datasets in deep learning. However, image compression usually adopts the form of lossy compression, which will lose part of the image information. If the lost information is key information, it may affect learning performance. In this paper, we will analyze the effect of image compression on deep learning performance on CNN-based cat and dog classification. Through the experiment results, we conclude that the compression of images does not have a significant impact on the accuracy of deep learning.