• ETRI Journal
• /
• v.42 no.6
• /
• pp.899-911
• /
• 2020

The data in tables are accurate and rich in information, which facilitates the performance of information extraction and question answering (QA) tasks. TableQA, which is based on tables, solves problems by understanding the table structure and searching for answers to questions. In this paper, we introduce both novice and intermediate Korean TableQA tasks that involve deducing the answer to a question from structured tabular data and using it to build a question answering pair. To solve Korean TableQA tasks, we use S3-NET, which has shown a good performance in machine reading comprehension (MRC), and propose a method of converting structured tabular data into a record format suitable for MRC. Our experimental results show that the proposed method outperforms a baseline in both the novice task (exact match (EM) 96.48% and F1 97.06%) and intermediate task (EM 99.30% and F1 99.55%).

• Journal of the Korea Society of Computer and Information
• /
• v.23 no.7
• /
• pp.105-111
• /
• 2018

Last March, the world Go competition between AlphaGo, AI Go program developed by Google Deep Mind and professional Go player Lee Sedol has shown us that the 4th industrial revolution using AI has come close. Especially, there ar many system combined with AI hae been developing including program for researching legal information, system for expecting jurisdiction, and processing big data, there is saying that even AI legal person is ready for its appearance. As legal field is mostly based on text-based document, such characteristic makes it easier to adopt artificial intelligence technology. When a legal person receives a case, the first thing to do is searching for legal information and judical precedent, which is the one of the strength of AI. It is very difficult for a human being to utilize a flow of legal knowledge and figures by analyzing them but for AI, this is nothing but a simple job. The ability of AI searching for regulation, precedent, and literature related to legal issue is way over our expectation. AI is evaluated to be able to review 1 billion pages of legal document per second and many people agree that lot of legal job will be replaced by AI. Along with development of AI service, legal service is becoming more advanced and if it devotes to ethical solving of legal issues, which is the final goal, not only the legal field but also it will help to gain nation's trust. If nations start to trust the legal service, it would never be completely replaced by AI. What is more, if it keeps offering advanced, ethical, and quick legal service, value of law devoting to the society will increase and finally, will make contribution to the nation. In this time where we have to compete with AI, we should try hard to increase value of traditional legal service provided by human. In the future, priority of good legal person will be his/her ability to use AI. The only field left to human will be understanding and recovering emotion of human caused by legal problem, which cannot be done by AI's controlling function. Then, what would be the attitude of legal people in this period? It would be to learn the new technology and applying in the field rather than going against it, this will be the way to survive in this new AI period.

• Transactions of the Korean hydrogen and new energy society
• /
• v.17 no.1
• /
• pp.31-38
• /
• 2006

There are two types of metal hydride electrodes as a negative electrode in a Ni-MH battery, $AB_2$ Zr-based Laves phases and $AB_5$ LM(La-rich mischmetal)-based alloys. The $AB_5$ alloy electrodes have characteristic properties such as a large discharge capacity per volume, easiness in activation, long cycle life and a low cost of alloy. However they have a relatively small discharge capacity per weight. The $AB_2$alloy electrodes have a much higher discharge capacity per weight than $AB_5$ alloy electrodes, however they have some disadvantages of poor activation behavior and cycle life. Therefore, in order to improve the discharge capacity of the $AB_5$ alloy electrode the Zr, Ti and V which are the alloying elements of the $AB_2$ alloys were added to the $LaNi_{3.6}Ai_{0.4}Co_{0.7}Mn_{0.3}$ alloy which was chosen as a $AB_5$ alloy with a high capacity. The addition of Zr, Ti and V to $LaNi_{3.6}Ai_{0.4}Co_{0.7}Mn_{0.3}$ alloy improved the activation to be completed in two cycles. The discharge capacities of Zr 0.02, Ti 0.02 and V 0.1 alloys in $LaNi_{3.6}Ai_{0.4}Co_{0.7}Mn_{0.3}M_y$ (M = Zr, Ti, V) were respectively 346, 348 and 366 mAh/g alloy. The alloy electrodes, Zr 0.02, Ti 0.05 and V 0.1 in $LaNi_{3.6}Ai_{0.4}Co_{0.7}Mn_{0.3}M_y$ (M = Zr, Ti, V), have shown good cycle property after 200 cycles. The rate capability of the $LaNi_{3.6}Ai_{0.4}Co_{0.7}Mn_{0.3}M_y$ (M = Zr, Ti, V) alloy electrodes were very good until 0.6 C rate and the alloys, Zr 0.02, Ti 0.05 and V 0.1, have shown the best result as 92 % at 2.4 C rate. The charge retention property of the $LaNi_{3.6}Ai_{0.4}Co_{0.7}Mn_{0.3}M_y$ (M = Zr, Ti, V) alloys was not good and the alloys with M content from 0.02 to 0.05 showed better charge retention properties.

• Proceedings of the Korean Information Science Society Conference
• /
• 2012.06b
• /
• pp.522-524
• /
• 2012

For exact classification of the defect, good feature selection and classifier is necessary. In this paper, various features such as brightness features, shape features and statistical features are stated and Bayes classifier using Gaussian mixture model is used as classifier. Also feature extraction method based on spectral graph theory is presented. Experimental result shows that feature extraction method using graph Laplacian result in better performance than the result using PCA.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.5
• /
• pp.1466-1488
• /
• 2022

Recently, the importance and necessity of artificial intelligence (AI), especially machine learning, has been emphasized. In fact, studies are actively underway to solve complex and challenging problems through the use of AI systems, such as intelligent CCTVs, intelligent AI security systems, and AI surgical robots. Information security that involves analysis and response to security vulnerabilities of software is no exception to this and is recognized as one of the fields wherein significant results are expected when AI is applied. This is because the frequency of malware incidents is gradually increasing, and the available security technologies are limited with regard to the use of software security experts or source code analysis tools. We conducted a study on MalDC, a technique that converts malware into images using machine learning, MalDC showed good performance and was able to analyze and classify different types of malware. MalDC applies a preprocessing step to minimize the noise generated in the image conversion process and employs an image augmentation technique to reinforce the insufficient dataset, thus improving the accuracy of the malware classification. To verify the feasibility of our method, we tested the malware classification technique used by MalDC on a dataset provided by Microsoft and malware data collected by the Korea Internet & Security Agency (KISA). Consequently, an accuracy of 97% was achieved.

• 한국정보교육학회:학술대회논문집
• /
• 2021.08a
• /
• pp.59-67
• /
• 2021

The development of AI technology has brought about a big change in our lives. As AI's influence grows from life to society to the economy, the importance of education on AI and data is also growing. In particular, the OECD Education Research Report and various domestic information and curriculum studies address data literacy and present it as an essential competency. Looking at domestic and international studies, one can see that the definition of data literacy differs in its specific content and scope from researchers to researchers. Thus, the definition of major research related to data literacy was analyzed from various angles and derived from various angles. In key studies, Word2vec natural language processing methods, along with word frequency analysis used to define data literacy, are used to analyze semantic similarities and nominate them based on content elements of curriculum research to derive the definition of 'understanding and using data to process information'. Based on the definition of data literacy derived from this study, we hope that the contents will be revised and supplemented, and more research will be conducted to provide a good foundation for educational research that develops students' future capabilities.

• Proceedings of the KSRS Conference
• /
• v.1
• /
• pp.236-238
• /
• 2006

An algorithm for detection of yellow sand aerosols has been developed with infrared bands from Moderate Resolution Imaging Spectroradiometer (MODIS) and Multi-functional Transport Satellite-1 Replacement (MTSAT-1R) data. The algorithm is the hybrid algorithm that has used two methods combined together. The first method used the differential absorption in brightness temperature difference between $11{\mu}m$ and $12{\mu}m$ (BTD1). The radiation at 11 ${\mu}m$ is absorbed more than at 12 ${\mu}m$ when yellow sand is loaded in the atmosphere, whereas it will be the other way around when cloud is present. The second method uses the brightness temperature difference between $3.7{\mu}m$ and $11{\mu}m$ (BTD2). The technique would be most sensitive to dust loading during the day when the BTD2 is enhanced by reflection of $3.7{\mu}m$ solar radiation. We have applied the three methods to MTSAT-1R for derivation of the yellow sand dust and in conjunction with the Principle Component Analysis (PCA), a form of eigenvector statistical analysis. As produced Principle Component Image (PCI) through the PCA is the correlation between BTD1 and BTD2, errors of about 10% that have a low correlation are eliminated for aerosol detection. For the region of aerosol detection, aerosol index (AI) is produced to the scale of BTD1 and BTD2 values over land and ocean respectively. AI shows better results for yellow sand detection in comparison with the results from individual method. The comparison between AI and OMI aerosol index (AI) shows remarkable good correlations during daytime and relatively good correlations over the land.

• Journal of Intelligence and Information Systems
• /
• v.29 no.3
• /
• pp.167-183
• /
• 2023

In the recent field of recommendation systems, various studies have been conducted to model sparse data effectively. Among these, GLocal-K(Global and Local Kernels for Recommender Systems) is a research endeavor combining global and local kernels to provide personalized recommendations by considering global data patterns and individual user characteristics. However, due to its utilization of kernel tricks, GLocal-K exhibits diminished performance on highly sparse data and struggles to offer recommendations for new users or items due to the absence of side information. In this paper, to address these limitations of GLocal-K, we propose the GEase-K (Global and EASE kernels for Recommender Systems) model, incorporating the EASE(Embarrassingly Shallow Autoencoders for Sparse Data) model and leveraging side information. Initially, we substitute EASE for the local kernel in GLocal-K to enhance recommendation performance on highly sparse data. EASE, functioning as a simple linear operational structure, is an autoencoder that performs highly on extremely sparse data through regularization and learning item similarity. Additionally, we utilize side information to alleviate the cold-start problem. We enhance the understanding of user-item similarities by employing a conditional autoencoder structure during the training process to incorporate side information. In conclusion, GEase-K demonstrates resilience in highly sparse data and cold-start situations by combining linear and nonlinear structures and utilizing side information. Experimental results show that GEase-K outperforms GLocal-K based on the RMSE and MAE metrics on the highly sparse GoodReads and ModCloth datasets. Furthermore, in cold-start experiments divided into four groups using the GoodReads and ModCloth datasets, GEase-K denotes superior performance compared to GLocal-K.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.372-376
• /
• 2023

본 연구에서는 비디오 하이라이트 검출 및 장면 추출을 위한 경량화된 모델인 Visual Context Learner (VCL)을 제안한다. 기존 연구에서는 매개변수가 고정된 CLIP을 비롯한 여러 피쳐 추출기에 학습 가능한 DETR과 같은 트랜스포머를 이어붙여서 학습을 한다. 하지만 본 연구는 경량화된 구조로 하이라이트 검출 성능을 개선시킬 수 있음을 보인다. 그리고 해당 형태로 장면 추출도 가능함을 보이며 장면 추출의 추가 연구 가능성을 시사한다. VCL은 매개변수가 고정된 CLIP에 학습가능한 프롬프트와 MLP로 하이라이트 검출과 장면 추출을 진행한다. 총 2,141개의 학습가능한 매개변수를 사용하여 하이라이트 검출의 HIT@1(>=Very Good) 성능을 기존 CLIP보다 2.71% 개선된 성능과 최소한의 장면 추출 성능을 보인다.

• Journal of The Korean Association of Information Education
• /
• v.25 no.5
• /
• pp.691-704
• /
• 2021

The development of AI technology has brought about a big change in our lives. The importance of AI and data education is also growing as AI's influence from life to society to the economy grows. In response, the OECD Education Research Report and various domestic information and curriculum studies deal with data literacy and present it as an essential competency. However, the definition of data literacy and the content and scope of the components vary among researchers. Thus, we analyze the semantic similarity of words through Word2Vec deep learning natural language processing methods along with the definitions of key data literacy studies and analysis of word frequency utilized in components, to present objective and comprehensive definition and components. It was revised and supplemented by expert review, and we defined data literacy as the 'basic ability of knowledge construction and communication to collect, analyze, and use data and process it as information for problem solving'. Furthermore we propose the components of each category of knowledge, skills, values and attitudes. We hope that the definition and components of data literacy derived from this study will serve as a good foundation for the systematization and education research of AI education related to students' future competency.