• Journal of Korea Society of Industrial Information Systems
• /
• v.29 no.3
• /
• pp.79-87
• /
• 2024

Recently, the importance of pre-trained language models (PLM) has been emphasized for natural language processing (NLP) such as text classification, sentiment analysis, and question answering. Korean PLM shows high performance in NLP in general-purpose domains, but is weak in domains such as finance, medicine, and law. The main goal of this study is to propose a language model learning process and method to build a financial-specific language model that shows good performance not only in the financial domain but also in general-purpose domains. The five steps of the financial-specific language model are (1) financial data collection and preprocessing, (2) selection of model architecture such as PLM or foundation model, (3) domain data learning and instruction tuning, (4) model verification and evaluation, and (5) model deployment and utilization. Through this, a method for constructing pre-learning data that takes advantage of the characteristics of the financial domain and an efficient LLM training method, adaptive learning and instruction tuning techniques, were presented.

• Journal of Korea Multimedia Society
• /
• v.10 no.12
• /
• pp.1752-1761
• /
• 2007

Location-based services are a recent concept that integrates a mobile device's location with other information in order to provide added value to a user. Although Location-based Services provide users with comfortable information, it is a complex task to manage and share heterogeneous and numerous data in decentralized environments. In this paper, we propose the Semantic LBS Model as one of the solution to resolve the problem. The Semantic LBS Model is a LBS middleware model that includes an ontology-based data model for LBS POI information and its processing mechanism based on Semantic Web technologies. Our model enables POI information to be described and retrieved over various domain-specific ontologies based on our proposed POIDL ontology. This mechanism provide rich expressiveness, interoperability, flexibility in describing and using information about POls, and it can enhance POI retrieval services.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 1991.10a
• /
• pp.469-474
• /
• 1991

Recent interests in intelligent LP modeling aim to support MS/OR-naive users to be able to apply LP models to practical problems without the expert knowledges required. For more generalized LP modeling, a GPS(General Problem Solver)-based approach is suggested in this paper. It identifies modeling process as a means-ends analysis process. In view of this approach, a) we first divide the knowledges into domain specific assertive knowledges(state) and procedural knowledges about LP modeling(operator and macro) for model-domain independence, b) and then generate LP model according to the difference resolution techniques.

• Journal of Information Processing Systems
• /
• v.17 no.5
• /
• pp.867-878
• /
• 2021

In context awareness and user intention tasks, dataset construction is expensive because specific domain data are required. Although pretraining with a large corpus can effectively resolve the issue of lack of data, it ignores domain knowledge. Herein, we concentrate on data domain knowledge while addressing data scarcity and accordingly propose a multi-channel long short-term memory (LSTM). Because multi-channel LSTM integrates pretrained vectors such as task and general knowledge, it effectively prevents catastrophic forgetting between vectors of task and general knowledge to represent the context as a set of features. To evaluate the proposed model with reference to the baseline model, which is a single-channel LSTM, we performed two tasks: voice phishing with context awareness and movie review sentiment classification. The results verified that multi-channel LSTM outperforms single-channel LSTM in both tasks. We further experimented on different multi-channel LSTMs depending on the domain and data size of general knowledge in the model and confirmed that the effect of multi-channel LSTM integrating the two types of knowledge from downstream task data and raw data to overcome the lack of data.

• IE interfaces
• /
• v.16 no.spc
• /
• pp.14-20
• /
• 2003

This paper presents a multi-criteria decision model and Multi-Criteria Generic Algorithm(MCGA) approach to improve backbone topology by leveraging the Virtual Link(VL) system in an hierarchical Link-State(LS) routing domain. Given that the sound backbone topology structure has a great impact on the overall routing performance in an hierarchical LS domain, the importance of this research is evident. The proposed decision model is to find an optimal configuration of VLs that properly meets two-pronged engineering goals in installing and maintaining VLs: i.e., operational costs and network reliability. The experiment results clearly indicates that it is essential to the effective operations of hierarchical LS routing domain to consider not only engineering aspects but also specific benefits from systematical layout of VLs, thereby presenting the validity of the decision model and MCGA.

• International Journal of Naval Architecture and Ocean Engineering
• /
• v.13 no.1
• /
• pp.340-350
• /
• 2021

This paper proposes a decision support model for USVs to improve the accuracy of collision avoidance decision-making. It is formed by Navigation Safety Domain (NSD) and domain-based Collision Risk Index (CRI), capable of determining the collision stage and risk between multiple ships. The NSD is composed of a warning domain and a forbidden domain, which is constructed under the constraints of COLREGs (International Regulations for Preventing Collisions at Sea). The proposed domain based CRI takes the radius of NSD in various encounter situations as threshold parameters. It is found that the value of collision risk in any directions can be calculated, including actual value and risk threshold. A catamaran USV and 6 given vessels are taken as study objects to validate the proposed model. It is found that the judgment of collision stage is accurate and the azimuth range of risk exists can be detected, hence the ships can take direct and effective collision avoidance measures. According to the relation between the actual value of CRI and risk threshold, the decision support rules are summarized, and the specific terms of COLREGs to be followed in each encounter situation are given.

• Convergence Security Journal
• /
• v.13 no.4
• /
• pp.11-17
• /
• 2013

All u-Health system has security vulnerabilities. This vulnerability locally(local) or network(network) is on the potential risk. Smart environment of health information technology, Ad-hoc networking, wireless communication environments, u-health are major factor to increase the security vulnerability. u-health care information systems user terminal domain interval, interval public network infrastructure, networking section, the intranet are divided into sections. Health information systems by separating domain specific reason to assess vulnerability vulnerability countermeasure for each domain are different. u-Healthcare System 5 layers of security risk assessment system for domain-specific security vulnerability diagnosis system designed to take the security measures are needed. If you use this proposed model that has been conducted so far vaguely USN-based health information network security vulnerabilities diagnostic measures can be done more systematically provide a model.

• ETRI Journal
• /
• v.27 no.2
• /
• pp.227-230
• /
• 2005

A specific anthropomorphic mannequin (SAM) model was used to investigate the relation between local specific absorption rate (SAR) and head size. The model was scaled to 80 to 100% sized models at intervals of 5%. We assumed that the shell of the SAM model has the same properties as the head-equivalent tissue. Five handsets with a monopole antenna operating at 835 MHz were placed in the approximate cheek position against the scaled SAM models. The handsets had different antenna lengths, antenna positions, body sizes, and external materials. SAR distributions in the scaled SAM models were computed using the finite-difference time-domain method. We found that a larger head causes a distinct increase in the spatial peak 1-voxel SAR, while head size did not significantly change the peak 1-g averaged-SAR and 10-g averaged-SAR values for the same power level delivered to the antenna.

• The KIPS Transactions:PartA
• /
• v.12A no.5 s.95
• /
• pp.451-460
• /
• 2005

We present a methodology to design energy-efficient data paths using FPGAs. Our methodology integrates domain specific modeling, coarse-grained performance evaluation, design space exploration, and low-level simulation to understand the tradeoffs between energy, latency, and area. The domain specific modeling technique defines a high-level model by identifying various components and parameters specific to a domain that affect the system-wide energy dissipation. A domain is a family of architectures and corresponding algorithms for a given application kernel. The high-level model also consists of functions for estimating energy, latency, and area that facilitate tradeoff analysis. Design space exploration(DSE) analyzes the design space defined by the domain and selects a set of designs. Low-level simulations are used for accurate performance estimation for the designs selected by the DSE and also for final design selection We illustrate our methodology using a family of architectures and algorithms for matrix multiplication. The designs identified by our methodology demonstrate tradeoffs among energy, latency, and area. We compare our designs with a vendor specified matrix multiplication kernel to demonstrate the effectiveness of our methodology. To illustrate the effectiveness of our methodology, we used average power density(E/AT), energy/(area x latency), as themetric for comparison. For various problem sizes, designs obtained using our methodology are on average $25\%$ superior with respect to the E/AT performance metric, compared with the state-of-the-art designs by Xilinx. We also discuss the implementation of our methodology using the MILAN framework.

• Journal of Internet Computing and Services
• /
• v.10 no.4
• /
• pp.139-149
• /
• 2009

In order to verify that a lot of legal requirements and regulations are correctly translated into software, this paper provides a solution for formal and computable representations of rules and requirements in data protection legislations with a DSML (Domain Specific Modeling Language). All policies are formally specified through Prolog and then integrated with DSML, According to the time of policy verification, this solution has two kinds of policies: static policies, dynamic policies.