• 지능정보연구
• /
• 제24권4호
• /
• pp.85-110
• /
• 2018

챗봇은 음성, 이미지, 비디오 또는 텍스트와 같은 다양한 매채를 이용하여 대화가 가능한 대화형 어시스턴트이자 인공지능을 기반으로 사용자의 질문에 답하거나 문제를 해결할 수 있는 사용자 친화적 프로그램이다. 하지만 현재 챗봇은 사용자가 요청한 작업을 정확하게 수행하는 기술적측면에 초점이 맞추어져 있으며, 개인화된 대화로 사용자와 챗봇간의 관계성 구축에는 제한적이어서 일부 사례에도 불구하고 소셜챗봇이 되기에는 미흡한 상태이다. 만약 인간의 사회성을 나타내는 특징 중 하나인 관계성을 챗봇이 인식하여 알맞게 대화를 하여 문제를 해결할 수 있다면, 개인화된 대화를 할 수 있을 뿐만 아니라 인간과 유사한 대화를 할 수 있을 것이다. 본 연구의 목적은 사용자가 입력한 내용을 기반으로 챗봇과 사용자 간의 관계성을 추론하고 대화 상황에 맞게 대화 상대가 적절한 대화를 수행 할 수 있는 텍스트 분석 방법을 제안하는 것이다. 본 연구의 실험 및 평가를 하기 위하여 실제 SNS대화 내용을 사용하였다. 분석결과 개인정보 보호를 위해 사용자의 개인 프로필 정보가 제외된 방법에서도 우수한 결과를 나타내어 소셜 챗봇에 적합한 방법으로 검증되었다.

• 융합정보논문지
• /
• 제10권12호
• /
• pp.9-14
• /
• 2020

현재 NTIS(National Technology Information Service)는 인공지능 기술을 기반으로 대화형 검색 서비스를 구축하고 있다. 이용자의 검색 의도를 파악하고 과제정보를 제공하기 위해 딥러닝 모델과 형태소 분석기를 기반으로 대화형 검색 서비스를 구축한다. 딥러닝 모델은 NTIS와 대화형 검색 서비스를 활용할 때 적재되는 로그 데이터를 기반으로 학습을 진행하고 이용자의 검색 의도를 파악한다. 그리고 단계별 검색을 통해 과제정보를 제공한다. 검색 의도 파악은 예외처리를 용이하게 해주며 단계별 검색은 통합검색보다 쉽고 빠르게 원하는 정보를 얻을 수 있도록 한다. 향후연구로는 인공지능 기술이 접목된 성장형 대화형 검색 서비스로써 이용자에게 제공하는 정보의 범위를 확대해야 한다.

• 신재생에너지
• /
• 제16권1호
• /
• pp.15-24
• /
• 2020

For mechanical load testing of wind turbines, capture matrix is constructed for various range of wind speeds according to the international standard IEC 61400-13. The conventional method wastes considerable amount of data by its invalid data policy -segment data into 10 minutes then remove invalid ones. Previously, we have suggested an alternative way to save the total amount of data to build a capture matrix, but the efficient selection of data has been still under question. The paper introduces optimization algorithms to construct capture matrix with less data. Heuristic algorithm (simple stacking and lowest frequency first), population method (particle swarm optimization) and Q-Learning accompanied with epsilon-greedy exploration are compared. All algorithms show better performance than the conventional way, where the distribution of enhancement was quite diverse. Among the algorithms, the best performance was achieved by heuristic method (lowest frequency first), and similarly by particle swarm optimization: Approximately 28% of data reduction in average and more than 40% in maximum. On the other hand, unexpectedly, the worst performance was achieved by Q-Learning, which was a promising candidate at the beginning. This study is helpful for not only wind turbine evaluation particularly the viewpoint of cost, but also understanding nature of wind speed data.

• Communications for Statistical Applications and Methods
• /
• 제27권3호
• /
• pp.313-326
• /
• 2020

Deep Learning is the most important key to the development of Artificial Intelligence (AI). There are several distinguishable architectures of neural networks such as MLP, CNN, and RNN. Among them, we try to understand one of the main architectures called Recurrent Neural Network (RNN) that differs from other networks in handling sequential data, including time series and texts. As one of the main tasks recently in Natural Language Processing (NLP), we consider Neural Machine Translation (NMT) using RNNs. We also summarize fundamental structures of the recurrent networks, and some topics of representing natural words to reasonable numeric vectors. We organize topics to understand estimation procedures from representing input source sequences to predict target translated sequences. In addition, we apply multiple translation models with Gated Recurrent Unites (GRUs) in Keras on English-Korean sentences that contain about 26,000 pairwise sequences in total from two different corpora, colloquialism and news. We verified some crucial factors that influence the quality of training. We found that loss decreases with more recurrent dimensions and using bidirectional RNN in the encoder when dealing with short sequences. We also computed BLEU scores which are the main measures of the translation performance, and compared them with the score from Google Translate using the same test sentences. We sum up some difficulties when training a proper translation model as well as dealing with Korean language. The use of Keras in Python for overall tasks from processing raw texts to evaluating the translation model also allows us to include some useful functions and vocabulary libraries as well.

• 한국시뮬레이션학회논문지
• /
• 제29권2호
• /
• pp.35-47
• /
• 2020

인구수 감소와 국방개혁으로 인한 병력 감축, 4차 산업혁명 기술의 초고도화에 따른 기술적 요인으로 국방 M&S의 개체를 자동화 모의하는 것은 이제 군의 현실적인 목표가 되었다. 자동화 모의 기술의 사용자인 군과 기술을 개발해야 하는 공학자들의 공통된 방향설정이 필요한 시점이다. 본 연구는 향후 국방 M&S의 자동화 모의 연구를 위한 가이드라인을 제시한다. 이를 위해, 먼저 자동화 모의를 가능하게 방법론들 규칙-기반방법, 에이전트-기반방법, 학습-기반방법에 대해 논의하고, 이어서 이러한 방법론을 어떠한 방향으로 개발해야 하는지에 대해 논의한다. 연구를 통해 국방 M&S의 자동화 모의 기술 연구가 본격화 되고, 군과 개발자 사이의 간극이 좁혀지기를 기대한다.

• 전자통신동향분석
• /
• 제35권7호
• /
• pp.46-55
• /
• 2020

As artificial intelligence (AI) technology becomes more important, the demand for AI talent is increasing. However, there is a shortage of AI talent around the world, and it is difficult to secure. Therefore, it has become more important to nurture the AI workforce. The private sector and government in Korea and other countries are making an effort to cultivate AI talent, and ETRI has proposed "Nurturing AI Professionals" as ETRI AI Strategy #5 to meet both internal and national demands for AI talent. ETRI has suggested three key tasks to implement AI Strategy #5. The first one is to create a "top-notch AI talent training project: the ETRI AI Academy" to strengthen AI research capabilities. The second one is "nurturing AI engineers specialized in local-based industries: the ETRI AI Business School" to help supply the necessary AI workforce in the industry. The third one is the "contribution to AI education service for people: ETRI AI Literacy" to raise the public's understanding and utilization of AI.

• Applied Microscopy
• /
• 제46권4호
• /
• pp.184-187
• /
• 2016

Neuronal connectivity determines brain function. Therefore, understanding the full map of brain connectivity with functional annotations is one of the most desirable but challenging tasks in science. Current methods to achieve this goal are limited by the resolution of imaging tools and the field of view. Macroscale imaging tools (e.g., magnetic resonance imaging, diffusion tensor images, and positron emission tomography) are suitable for large-volume analysis, and the resolution of these methodologies is being improved by developing hardware and software systems. Microscale tools (e.g., serial electron microscopy and array tomography), on the other hand, are evolving to efficiently stack small volumes to expand the dimension of analysis. The advent of mesoscale tools (e.g., tissue clearing and single plane ilumination microscopy super-resolution imaging) has greatly contributed to filling in the gaps between macroscale and microscale data. To achieve anatomical maps with gene expression and neural connection tags as multimodal information hubs, much work on information analysis and processing is yet required. Once images are obtained, digitized, and cumulated, these large amounts of information should be analyzed with information processing tools. With this in mind, post-imaging processing with the aid of many advanced information processing tools (e.g., artificial intelligence-based image processing) is set to explode in the near future, and with that, anatomic problems will be transformed into informatics problems.

• 한국가스학회지
• /
• 제8권2호
• /
• pp.28-34
• /
• 2004

Enhanced methodologies for process diagnosis and abnormal situation management have been developed for the last two decades. However, there is no single method that always shows better performance over all kinds of diagnostic problems. In this paper, a framework of message-passing, cooperative, intelligent diagnostic agents is presented for improved on-line fault diagnosis through cooperative problem solving of different expertise. A group of diagnostic agents in charge of different process functional perform local diagnoses in parallel; exchange related information with other diagnostic agents; and cooperatively solve the global diagnostic problem of the whole process plant or business units just like human experts would do. For their better understanding, sharing and exchanging of process knowledge and information, we also suggest a way of remodeling processes and protocols, taking into account semantic abstracts of process information and data. The benefits of the suggested multi-agents-based approach are demonstrated by the implementations for solving the diagnostic problems of various chemical processes.

• Nuclear Engineering and Technology
• /
• 제53권3호
• /
• pp.776-792
• /
• 2021

Conventional integral effect test facilities were constructed to enable the precise observation of thermal-hydraulic phenomena and reactor behaviors under postulated accident conditions to prove reactor safety. Although these facilities improved the understanding of thermal-hydraulic phenomena and reactor safety, applications of new technologies and their performance tests have been limited owing to the cost and large scale of the facilities. Various nuclear technologies converging 4th industrial revolution technologies such as artificial intelligence, drone, and 3D printing, are being developed to improve plant management strategies. Additionally, new conceptual passive safety systems are being developed to enhance reactor safety. A new integral effect test facility having a noticeable scaling ratio, i.e., the (UNIST reactor innovation loop (URI-LO), is designed and constructed to improve the technical quality of these technologies by performance and feasibility tests. In particular, the URI-LO, which is constructed using a transparent material, enables better visualization and provides physical insights on multidimensional phenomena inside the reactor system. The facility design based on three-level approach is qualitatively validated with preliminary analyses, and its functionality as a test facility is confirmed through a series of experiments. The design feature, design validation, functionality test, and future utilization of the URI-LO are introduced.

• 전자통신동향분석
• /
• 제33권6호
• /
• pp.58-68
• /
• 2018

A neuromorphic hardware that mimics biological perceptions and has a path toward human-level artificial intelligence (AI) was developed. In contrast with software-based AI using a conventional Von Neumann computer architecture, neuromorphic hardware-based AI has a power-efficient operation with simultaneous memorization and calculation, which is the operation method of the human brain. For an ideal neuromorphic device similar to the human brain, many technical huddles should be overcome; for example, new materials and structures for the synapses and neurons, an ultra-high density integration process, and neuromorphic modeling should be developed, and a better biological understanding of learning, memory, and cognition of the brain should be achieved. In this paper, studies attempting to overcome the limitations of next-generation neuromorphic hardware technologies are reviewed.