Earticle

이 연구는 국내에서 진행되는 한국어 교육의 상황에서 한국어 학습자들 대상으로 번역 수업을 실현하기 위한 한국어 교수-학습 방법을 제시한 것이다. 이를 위해 먼저 한국어교육 분야에서 번역과 관련된 선행 연구를 검토했다. 그리고 번역을 어떻게 효과적으로 교육할 수 있는지를 파악하기 위해 번역에 대한 이론적 개념을 살펴보았으며, Pacte의 연구에 기초하여 번역 능력의 구성요소를 살펴보았다. 그리고 외국어 교육에서 활용되는 TTT 모형을 확대하여 한국어 번역 수업에 적용할 수 있는 ‘과제중심 번역 교수 모형’을 구안했다. 한국어 번역 수업의 실제를 보이는 부분에서는 ‘한국어 번역’ 교과목의 구성을 설계한 후에, 실제로 ‘과제중심 번역 교수 모형’을 적용하여 한국어 번역 수업이 이루어지는 과정을 제시했다. 그리고 번역 수업 과정과 번역물을 평가하는 방법을 논의했다. 이 연구는 국내외에서 한국어 번역 교육에 관심이 증가하고 있는 상황에서 교실 수업에서 실행이 가능한 번역 교수-학습 방법을 구안하여 제시하였다는 점에서 의의가 있다.

This study presents a Korean language teaching–learning method for translation classes for Korean-language learners in Korea. To this end, the researcher first reviewed previous studies related to translation in the field of Korean language education. The theoretical concept of translation and the components of translation competence based on Pacte’s research were examined in order to understand how to effectively teach translation. In addition, the TTT model used in foreign language education was expanded to devise a “task-based translation teaching model” that could be applied to Korean translation classes. To demonstrate the reality of Korean translation classes, a syllabus for a “Korean translation” course was designed and the process of actually applying the translation teaching model is suggested. Furthermore, the author presents a method for evaluating translations. This study is significant in that it devised and suggested a translation teaching–learning method that can be implemented in classroom instruction in the context of increasing interest in Korean translation education in Korea and abroad.

연구목적: 본 연구는 플랜트 건설현장에서 유해·위험요인을 체계적으로 식별하고 평가할 수 있는 딥러닝 기 반 작업 위험성평가 모델을 구축하는 데 목적이 있다. 특히, 경험과 지식이 부족한 근로자도 쉽게 활용할 수 있도록 체계적인 시스템을 설계하여, 기존의 경험기반 평가 방식의 문제를 해결하고자 한다. 연구방법: 국내 플랜트 건설사인 A사의 976,140건의 위험성평가 데이터를 기반으로 데이터 정제 및 전처리를 수행하고, LDA 기반 토픽 모델링과 심층 신경망(DNN)을 활용하여 위험성 평가 모델을 구축하였다. 데이터셋 18,000 개를 사용하여 학습 및 테스트를 진행했으며, 최종 검증은 1,026건의 신규 위험성평가 데이터를 활용하였다. 연구결과: 개발된 모델은 작업조건에 따라 유해·위험 요인과 재해 유형 등을 체계적으로 추천하였고, 추천 결 과의 95.6% 이상이 전문가에 의해 적합하다고 평가되었다. 최종검증을 통해 8개 공종의 위험성평가 1,026개 에 대해 추천된 6,380개 전체가 추천 적합도 80 이상으로 모델 설계 기준을 만족하였고, 전 공종에서 일관된 추천 품질과 높은 적합성을 지님을 입증하였다. 결론: 대규모 비정형 안전 데이터를 정형화하여 데이터 기반 의 체계적인 위험성평가 모델을 구축함으로써, 기존의 경험 및 주관적인 평가 방식에 한계를 보완하였다. AI 기술을 활용을 통해 위험 요소 도출의 일관성과 효율성을 제고하고 위험성 평가의 실효성을 높였다.

Purpose: This study aims to develop a deep learning-based task risk assessment model capable of systematically identifying and evaluating hazardous and risk factors at plant construction sites. The model is specifically designed systematically to be easily utilized by workers with limited experience or knowledge, thereby addressing the issues associated with traditional experience-based assessment methods. Method: A total of 976,140 risk assessment records from domestic construction Company A. were cleansed and pre-processed. A risk assessment model was then constructed using LDA-based topic modeling and a deep neural network(DNN). The model was trained and tested using a dataset of 18,000, with final validation conducted on 1,026 new risk assessments. Result: The developed model systematically recommends hazardous and risky factors and types of accidents based on working conditions, with more than 95.6% of the recommendations evaluated as appropriate by experts. In the final validation, the model recommended 6,380 items for 1,026 assessments across eight work types, all achieving suitability score of above 80, thereby meeting the model design criteria and demonstrating consistent recommendation quality and high applicability across all construction domains. Conclusion: By structuring large-scale unstructured safety data into a data-driven risk assessment model, this study overcomes the limitations of traditional subjective evaluation methods. The application of AI enhances consistency and efficiency in hazard identification, improving the effectiveness of risk assessments.

대화시스템에서 자연어 생성은 대화관리 단계에서 결정한 시스템 발화의 의미표현을 사람이 이해할 수 있는 자연어 로 생성하는 것이다. 기존의 자연어 생성 연구는 의미표현에 대하여 매우 제한된 종류의 발화만을 생성하거나 문법 적으로 불완전한 발화를 생성한다는 문제점이 있다. 그래서 본 논문에서는 문제점들을 동시에 처리하기 위하여 Long Short Term Memory 기반의 언어모델을 이용한 한국어 자연어 생성 모델을 제안한다. 특히 우리는 시스템 발화의 다양성과 문법적 정확성을 높이기 위하여 빔서치 디코딩을 적용한다. 실험은 어절, 형태소, 음절단위에 따라 개별적으로 진행하였으며, 생성한 문장들은 정량적, 정성적 평가를 모두 진행하였다. 그 결과 형태소 단위로 학습한 제안모델에 빔서치 디코딩을 적용한 방법은 가장 좋은 성능을 보였다. 실제로 해당 생성 문장은 정량평가 결과에서 BLEU 지표는 0.86, Slot Error Rate 지표는 0.03을 기록하였으며 정성평가 역시 문법적으로 정확하고 문맥적 으로 충분히 자연스러운 결과임을 확인하였다.

Natural language generation in the dialogue system is a task that transforms the semantic frame of the system utterance determined in the dialogue management phase into a natural language that can be understood by humans. Existing studies have still faced some obstacles in that only very limited types of utterances or grammatically incomplete ones are generated from the semantic frames. In order to address these issues simultaneously, we propose a Korean natural language generation model using a long short term memory based language model. In particular, we exploit the beam search decoding method to obtain system utterances with diverse structures and grammatical correctness. The experiments were conducted individually with respect to the word, morpheme, and syllable units, and the generated utterances were evaluated in both quantitative and qualitative ways. As a result, the morpheme-based model with the beam search decoding has achieved the most robust result of all. In fact, in the quantitative evaluation result of the generated sentence, the BLEU-4 score was 0.86 and the SER was 0.03, and the qualitative evaluation was also confirmed to be grammatically correct and contextually natural.

본 연구는 Lavelle과 McMahan, Harris(2009)의 대상유사성 모형(target similarity model)에 기반해 종업원의 자원지각이 자긍심 발달을 거쳐 몰입을 야기하는 매개과정을 조직과 직무차원으로 각각 구분해서 살펴보았다. 대기업 종업원(746명)을 대상으로 매개모형을 분석한 결과, 조직차원과 직무차원 각각에서 대상유사성 효과에 해당하는 간접효과가 모두 유의한 것으로 나타났고, 누출효과(spillover effect)에 해당하는 교차경로를 포함한 매개효과는 모두 유의하지 않았다. 다음으로 간접효과 간 차이비교를 통해 지각된 조직지원이 조직기반 자긍심을 거칠 때 과업특정적 자긍심을 거치는 경우보다 효과가 유의하게 더 강한 것으로 나타났고, 직무차원에서는 유의한 차이를 발견할 수 없었다. 각 검증에서 구조방정식모형 프로그램인 MPLUS를 사용하여 Bootstraping을 통한 효과크기들의 신뢰구간을 추정하였으며 관련 시사점과 제한점을 논의하였다.

The purpose of this study was to examine the mediating role of two self-esteem (organization-based self-esteem: OBSE; task-specific self-esteem: TSSE) in predicting two types of employee commitment (job involvement, organizational commitment) based on the framework of Lavelle, McMahan and Harris(2009)’s target similarity model. A sample of 746 south korean employees were participated in this study and data were analyzed by MPLUS 6.12. The main results are as follows. First, the indirect effects reflecting target similarity effect were supported, but another indirect effects which reflect spillover effect between two dimensions was not. Second, the result of comparison between target similarity effect and spillover effect, which has same predictor and criterion but different mediator in each dimensions, was significant in organization domain, but not in job domain. Finally, the implications and limitations were discussed.

EA의 업무를 수행하기 위해 마련해 놓은 지식 체계인 업무 참조 모델(Business Reference Model)은 대부분 범주 별로 계층형 구조화 되어 있으며, 해당 조직의 목표를 달성하고 업무 수행능력 의 개선을 지원한다. 이를 통해 조직의 효율성을 높이고 조직 업무의 예외적 돌발성을 최소화하고 업 무간의 연계성 강화를 통해 개별 업무절차를 직관적이고 단순하게 변화시키고, 이로써 조직 내적으로 효율성 향상에 기여할 수 있다. EA에는 현재 BOK(Body of Knowledge)가 정의되어 존재하지만, 이 들을 업무의 할당과 그 평가에 사용하기 위한 시스템이 미비한 상태이다. 본 연구에서는 계층적 유사 도에 근거한 업무 할당 및 평가체계를 제공하고, 이를 통해 기존에는 정성적으로 이루어지던 업무의 정의와 할당 및 평가체계에 수치적 엄밀성을 도입하여 정량화된 명확한 평가가 가능하도록 하고, 이 것이 되먹임 되어 BOK체계 자체가 재평가되고 성숙될 수 있는 방법론을 제안하였다.

BRM (Business Reference Model) is constructed for enhancing the performance of the EA (Enterprise Architecture) tasks with respect to categories defined as a knowledge hierarchy. It aims to accomplish the organizational purposes by which it progresses enactment of the tasks. Furthermore, it simplifies the definition of the EA tasks and improves the effectiveness of them mainly due to reducing redundancy and reinforcing in-between connectivity of the tasks. By doing these, the internal organization performance can be exceedingly amended. In EA, there usually exists a hierarchically structured BOK (Body of Knowledge). The utilization of BOK such as the coordination for task assignment and the corresponding assessment cannot properly be consummated. In the paper, we introduce a distance measure for the categorical data based on the hierarchical knowledge so that the EA task assignment and assessment method can effectively prepared, and the BRM per se can be matured by the feedback of the BOK.

해군은 다양한 해상작전을 수행하기 위해 해상기동부대(TF)를 구성한다. 해양 환경하에서 해상기동부 대(TF)는 대함전(ASUW), 대잠전(ASW), 대공전(AAW), 그리고 상륙돌격 작전 등과 같은 동시·병 행적으로 성분작전을 수행한다. TF는 임무완수를 위하여 C4I, 인터넷 음성전화(VOIP) 및 디지털 전 보처리체계(DMHS)와 LINK-11, 16, ISDL 등과 같은 전술데이터링크를 갖춘 많은 전술체계들을 구 성하고 있다. TF가 수명한 임무를 완수하기 위해 해상작전을 수행할 때, TF의 작전절차에 관심을 갖 게 되었다. 작전절차는 적을 격파하기 위한 해상작전 수행을 위한 표준절차이다. 각 함정은 ‘어떻게 싸 울 것인가’에 대한 매뉴얼에 교전절차는 갖고 있지만, 현재 TF의 상세한 작전절차는 미흡하다. 따라 서, 본 논문에서 우리는 효과적인 해상작전 수행에 필요한 작전절차를 제안한다. 그런 작전절차 환경 에서 TF의 작전 효과성은 전술데이터링크 운영의 작전 시나리오를 통해 결정하고자 한다. 전술데이 터링크를 적용한 해상작전에서 데이터 링크 기반으로 국방 아키텍처프레임워크(MND-AF) OV-6c(운 용상태 추적 기술서)에서 작전정보 공유효과를 살펴보고, 정보 전파과정 개선을 식별하는 것이다. 본 논문에서는 효과적인 해군작전의 해상 TF를 위한 작전절차 모델을 제안하였다.

Navy establishes the Naval Task Forces (TF) for many kinds of maritime operations. Then the TF in the maritime environment performs simultaneous component operations such as ASUW (Anti-Surface Warfare), ASW (Anti-Submarine Warfare), AAW (Anti-Aircraft Warfare), and assault operations. The TF consists of many tactical systems for the completion of missions - C4I, VOIP (Voice Over Internet Protocol), DMHS (Digital Massage Handling System), and TDLs (Tactical Data Links) such as LINK-11, 16, ISDL (Inter Site Data Link). When the TF executes naval operations to complete a mission, we are interested in the kill chain for the maritime operations in the TF. The kill chain is a standard procedure for the naval operations to crush enemy defenses. Although each ship has a procedure about a manual for ‘how to fight’, it leave something to be desired for the TF detailed kill chain currently. Therefore, in this paper, we propose the naval TF's kill chain to perform the naval operations. Then, the operational effectiveness of the TF in the kill chain environment is determined through operation scenarios of TDL system implementation. It is to see the operational information sharing effect to a data link model based on MND-AF OV-6c (statement of tracking operational status) in the maritime operations applied to TDL and is to identify improvements in information dissemination process. We made the kill chain of maritime TF for the effective naval operations.

본 연구의 목적은 문헌정보학 연구 방법 분석의 맥락에서 과업 유형에 따른 대규모 언어모델(LLM) 기반 내용 분석의 적용 가능 조건을 차원별로 비교․검토하는 데 있다. 이를 위해 2020년부터 2024년까지 국내 4대 문헌정보학 학회지에 게재된 설문 및 인터뷰 연구 100편을 층화무작위표집 방식으로 선정하고, 표집 방법론을 구성하는 12개 차원에 대해 인간 코더 1인과 4개 대규모 언어모델(Claude-3.5-Haiku, GPT-4o-Mini, Gemini-2.0-Flash, Grok-4-Latest)의 코딩 결과를 비교하였다. 분석 결과, 명시적 기준에 따라 분류가 가능한 차원에서는 상대적으로 높은 일치도가 나타난 반면, 추론적․평가적 판단을 요구하는 차원에서는 일관되게 낮은 수준의 일치도가 확인되었다. 이러한 결과는 LLM 기반 자동화 코딩의 성과가 모델 성능 자체보다는 과업의 판단 구조와 정보의 명시성에 더 크게 영향을 받음을 시사한다. 따라서 LLM의 활용 범위는 과업 유형 및 판단 특성 차원에서 보다 정교하게 검토될 필요가 있으며, 인간-AI 혼합 검증 전략의 체계적 설계가 요구된다.

The purpose of this study is to compare and examine, across multiple dimensions, the conditions under which large language model (LLM)-based content analysis can be applied according to task type in the context of research methods analysis in library and information science. To this end, 100 survey and interview studies published between 2020 and 2024 in four major Korean journals in library and information science were selected using stratified random sampling. The coding results produced by one human coder and four large language models (Claude-3.5-Haiku, GPT-4o-Mini, Gemini-2.0-Flash, and Grok-4-Latest) were compared across twelve dimensions constituting sampling methodology. The results show that relatively high levels of agreement were observed in dimensions where classification could be made based on explicit criteria, whereas consistently lower levels of agreement appeared in dimensions requiring inferential or evaluative judgment. These findings suggest that the performance of LLM-based automated coding is influenced more by the decision structure of the task and the explicitness of the available information than by model performance itself. Therefore, the scope of LLM application should be more carefully examined from the perspectives of task type and judgment characteristics, and the systematic design of human-AI hybrid validation strategies is required.

With the recent advances in mobile platform technologies, a variety of studies on context-aware information services for the tourist information domain have been undertaken. Many studies on ontological approaches to tourist information services, moreover, have been conducted. However, most studies have focused on upper-level or domain ontologies; comparatively, few have proceeded from the perspective of task ontology based on mobile users’ generic tasks. Thus, we considered the construction of a task model and task ontology based on mobile users’ generic activities for a task-oriented tourist information service. In this paper, we introduce 1) a generic task model based on travelers’ needs and generic activities before and during trips, which model accounts for generic tasks and task processes; 2) a task ontology based on the generic task model, and 3) a task-ontology-based Task-Oriented Tourist Information Service (TOTIS). Using the generic task model and the task ontology, task-oriented menu can be constructed automatically by means simply users’ selections and context-awareness. Additionally, compared with the existing domain-oriented services, the TOTIS can facilitate more flexible searching of tourist information and make real-time determinations with context-awareness.

Mobile edge computing (MEC) plays a crucial role in improving the performance of resource-constrained mobile devices by offloading computation-intensive tasks to nearby edge servers. However, existing methods often neglect the critical consideration of future task requirements when making offloading decisions. In this paper, we propose an innovative approach that addresses this limitation. Our method leverages recurrent neural networks (RNNs) to predict task sizes for future time slots. Incorporating this predictive capability enables more informed offloading decisions that account for upcoming computational demands. We employ genetic algorithms (GAs) to fine-tune fitness functions for current and future time slots to optimize offloading decisions. Our objective is twofold: minimizing total processing time and reducing energy consumption. By considering future task requirements, our approach achieves more efficient resource utilization. We validate our method using a real-world dataset from Google-cluster. Experimental results demonstrate that our proposed approach outperforms baseline methods, highlighting its effectiveness in MEC systems.

The applications based on cloud computing platform usually need to use a number of computing resources and storage resources to completing computing tasks, so the fault-tolerant capability of system has become increasingly important. Aiming to solve this problem, an evaluation model of task scheduling is proposed based on cloud system (TSCS). TSCS can effectively model and simulate complex cloud systems due to its strong capabilities of quantitative evaluation and behavioral description resulting from combining the theoretical characteristics of queuing task theory and Petri net. The algorithm solves the problem that meeting customer service satisfaction and load balancing at the same time. In addition, consider single backup task status, for the failure of more than one processor at the same time, present the minimum cost of backup scheduling algorithm, the algorithm to solve the problem that require a lot of backup cost. Experimental results show that TSCS is able to effectively reflect the architecture characteristics of various cloud system at the perspectives of performance, service, etc., and highly simulated various kinds of dynamic service behaviors of cloud system, single workload and multi workloads shows that the proposed policy can finish the user’s queuing task scheduling before deadline as well as obtain approving cost efficient.

Fundamental information of the design unit is described using the correlation between the nodes in the task chain. Data transfer between design units is formulated as fundamental data transfer, design rule transfer and path scheme transfer, respectively. The design process is stored with the node as the unit by using the algorithm for decomposing correlated nodes. The reconstruction method is employed to eliminate the redundant nodes that exist in various previous design processes, alleviating the degree of coupling. The performance of the proposed scheme is verified by applying it to the development of the low-voltage appliance.

With the development of cloud computing technology, people not only want to pursue the shortest time to complete the tasks by using cloud computing, but also hope to take into the running costs of machines. Existing task scheduling algorithm in the cloud computing environment has been unable to meet people's needs. As an extension and generalization of the model checking theory, probability model checking is also used in many fields, such as random distributed algorithm and other areas. The task scheduling algorithm based on the particle swarm optimization algorithm combined with probability model is proposed in this paper. The algorithm defines the fitness functions of the time cost and the running cost. The fitness functions can improve the efficiency of the cloud computing platform. At the same time, the probability model can be used to analyze the running states of machines and the computing capability of the nodes in the cloud cluster. The probability, which is calculated by the probability model, provides the basis for changing particle swarm algorithm’s the inertia factor and the learning factor, so as to solve the drawback that the inertia factor and the learning factor solely depend on the fixed value.

We proposed a task scheduling in cloud computing based on intelligence firefly algorithm aimed at the disadvantages of cloud computing task scheduling. Firstly, on the basis of cloud model, used intelligence firefly algorithm with strong ability of global searching to find the better solution of cloud computing task scheduling then turned the better solution into the initial pheromone of improved firefly algorithm, and found out the cloud computing task scheduling and the algorithm’s global optimal solution through improved firefly information communications and feedbacks. Finally, made comparison test of the three benchmark function on the basis of MATLAB, the results showed, compared with traditional intelligence firefly algorithms, the improved algorithm can preferably allocate the resources in cloud computing model, the effect of prediction model time is more close to actual time, can efficiently limit the possibility of falling into local convergence, the optimal solution’s time of objective function value is shorten which meet the user’s needs more.

Federated learning provides an efficient integrated model for distributed data, allowing the local training of different data. Meanwhile, the goal of multi-task learning is to simultaneously establish models for multiple related tasks, and to obtain the underlying main structure. However, traditional federated multi-task learning models not only have strict requirements for the data distribution, but also demand large amounts of calculation and have slow convergence, which hindered their promotion in many fields. In our work, we apply the rank constraint on weight vectors of the multi-task learning model to adaptively adjust the task's similarity learning, according to the distribution of federal node data. The proposed model has a general framework for solving optimal solutions, which can be used to deal with various data types. Experiments show that our model has achieved the best results in different dataset. Notably, our model can still obtain stable results in datasets with large distribution differences. In addition, compared with traditional federated multi-task learning models, our algorithm is able to converge on a local optimal solution within limited training iterations.

This paper proposes a new model by combining an infinite-server queueing model for multi-task processing software system with a perfect debugging model based on Markov process with two types of faults suggested by Lee et al. (2001). We apply this model for module and integration testing in the testing process. Also, we compute several measure, such as the expected number of tasks whose processes can be completed and the task completion probability are investigated under the proposed model.

Amid the rapid evolution of digital health services, medical escort platforms that integrate healthcare delivery and digital technology are increasingly vital for enhancing the user experience and improving healthcare accessibility. To investigate the adoption mechanisms of such platforms, this study integrates Task?Technology Fit (TTF) theory and the Extended Technology Acceptance Model (ETAM) to develop a multi-path model of behavioral intention. Empirical validation was conducted using 522 valid samples collected in China. The results showed that Task?Technology Fit has a significant direct effect on usage intention and exerts notable indirect effects via cognitive variables, including perceived usefulness, ease of use, professionalism, and security. Perceived usefulness, as a core construct of ETAM, is a central mediating node across multiple paths, demonstrating chained and straightforward mediation mechanisms. Additionally, perceived professionalism and perceived security, introduced as context-specific extensions, significantly enhance users' perceived value and adoption intention, as influenced by service reliability and technological safety. These findings extended the theoretical boundaries and contextual applicability of Technology Acceptance Models within multitasking, privacy-intensive healthcare environments. This study revealed how task-technology fit influences user decision-making through multiple cognitive pathways. It deepens the explanatory scope of technology?cognition?behavior models. It also offers systematic insights for platform optimization, service design, regulatory governance, and inclusive digital healthcare.

This paper deals with providing a surgical robot with awareness of the current surgical stage. The awareness of the surgical stage is the first step toward a natural interaction between a surgeon and a surgical robot, the ultimate goal of which is to help the surgeon perform surgery with a minimum control burden. For this purpose, a surgery task model was defined as a structured form of surgical knowledge, which can be understood by both the surgeon and the robot. The model consists of three components: a surgical procedure model, input information, and an action strategy at each surgical stage. This paper focuses on the awareness of current surgical stages based on the surgical procedure model. The surgical procedure model represents the sequential information of the surgery and it is arranged based on key surgical stages. To implement the surgical procedure model of a cholecystectomy, 21 cases of human cholecystectomies are decomposed into surgical stages and their relations are then analyzed. To deal with uncertainty, interaction functions are introduced to the model. While further experiments are necessary, it was shown that the key stages-based surgical procedure model could estimate the key surgical stages correctly during one case of in vivo porcine cholecystectomy.