• Journal of Information Processing Systems
• /
• 제17권5호
• /
• 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.

• 한국컴퓨터정보학회논문지
• /
• 제24권1호
• /
• pp.57-65
• /
• 2019

In this paper, we propose a Rights Management Information(RMI) expression systems for individual sites are integrated and the performance evaluation is performed to find out an efficient comparing and updating method of RMI through various image feature point search techniques. In addition, we proposed a weighted scoring model for both public domain sites and posts in order to use the most latest RMI based on reliable data. To solve problem that most public domain sites are exposed to copyright infringement by providing inconsistent RMI(Rights Management Information) expression system and non-up-to-date RMI information. The weighted scoring model proposed in this paper makes it possible to use the latest RMI for duplicated images that have been verified through the performance evaluation experiments of SIFT and CNN techniques and to improve the accuracy when applied to search engines. In addition, there is an advantage in providing users with accurate original public domain images and their RMI from the search engine even when some modified public domain images are searched by users.

• 말소리와 음성과학
• /
• 제15권3호
• /
• pp.83-88
• /
• 2023

This study investigates the fine-tuning of large-scale Automatic Speech Recognition (ASR) models, specifically OpenAI's Whisper model, for domain-specific applications using the KsponSpeech dataset. The primary research questions address the effectiveness of targeted lexical item emphasis during fine-tuning, its impact on domain-specific performance, and whether the fine-tuned model can maintain generalization capabilities across different languages and environments. Experiments were conducted using two fine-tuning datasets: Set A, a small subset emphasizing specific lexical items, and Set B, consisting of the entire KsponSpeech dataset. Results showed that fine-tuning with targeted lexical items increased recognition accuracy and improved domain-specific performance, with generalization capabilities maintained when fine-tuned with a smaller dataset. For noisier environments, a trade-off between specificity and generalization capabilities was observed. This study highlights the potential of fine-tuning using minimal domain-specific data to achieve satisfactory results, emphasizing the importance of balancing specialization and generalization for ASR models. Future research could explore different fine-tuning strategies and novel technologies such as prompting to further enhance large-scale ASR models' domain-specific performance.

• 한국세라믹학회지
• /
• 제26권1호
• /
• pp.37-42
• /
• 1989

Periodic layered domain structures in doped LiNbO3 crystals grown by Czochralski method were obtained by thermal fluctuation and crystal rotation with inhomogeneous radial temprature distribution. The stressinduced domain formation mechanism model was suggested and discussed.

• 정보처리학회 논문지
• /
• 제13권2호
• /
• pp.18-25
• /
• 2024

이 연구는 도시 도메인 정보 모델 간의 상호운용성을 확보하기 위해 모델 레지스트리 활용을 제안하였다. 또한, 도시 정보와 데이터 공유, 관련된 표준 및 표준화 활동을 검토하고, 공간 데이터 인프라(SDI)에서의 레지스트리 사례를 분석하여 도시 도메인 정보 모델 간의 정보 공유와 교환을 위한 모델 레지스트리 개념을 제안하였다. 제안된 개념을 구체화하기 위해 끊김 없는 스마트시티 서비스 구현을 전제로 모델 레지스트리를 활용하여 도시 도메인 정보 모델을 연계하는 유스케이스를 개발하였다. 이를 기반으로 기술적 요구사항을 논의하고 프로토타입 개발을 통한 기술 구현과 실증을 향후 연구과제로 제시하였다.

• 대기
• /
• 제22권1호
• /
• pp.109-115
• /
• 2012

We parallelized the CFD_NIMR model, which is a numerical meteorological model, for best performance on both of distributed and shared memory parallel computers. This hybrid parallelization uses MPI (Message Passing Interface) to apply horizontal 2-dimensional sub-domain out of the 3-dimensional computing domain for distributed memory system, as well as uses OpenMP (Open Multi-Processing) to apply vertical 1-dimensional sub-domain for utilizing advantage of shared memory structure. We validated the parallel model with the original sequential model, and the parallel CFD_NIMR model shows efficient speedup on the distributed and shared memory system.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2001년도 하계학술대회 논문집 D
• /
• pp.2114-2116
• /
• 2001

Model reduction from high order systems to low order systems in frequency domain is considered four point (${\angle}$G(jw)=0, - ${\pi}/2$, ${\pi}$, and -3${\pi}$/2) instead of two point (${\angle}$G(jw) = - ${\pi}$/2,- ${\pi}$) of existing method in Nyquist curve. The Performances of reduced order model by proposed approach is similar to original model. In this paper, we proposed a new tuning algorithm for PID controller using model reduction in frequency domain. Simulations for some examples with varies dynamic characteristics are provided to show the effectiveness of the proposed tuning algorithm for PID controller using model reduction.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1996년도 하계학술대회 논문집 B
• /
• pp.731-733
• /
• 1996

In this paper, a new time-domain reduction method for unbalanced 3 phase power systems will be represented. The impulse response of the system is used to identify a discrete-time equivalent filter model. The model is formulated directly in the phase domain. Each phase has a self-mode equivalent model and two mutual-mode equivalent models. The equivalent model is determined by the transfer function identification technique based on the Prony analysis. The model is implemented in EMTDC and tested with an unbalanced 3 phase network. The result of test showed that the equivalent model is accurate.

• Advances in aircraft and spacecraft science
• /
• 제9권3호
• /
• pp.217-242
• /
• 2022

This paper proposes a novel time-domain homogenization model combining the viscoelastic constitutive law with Eshelby's inclusion theory-based micromechanics model to predict the mechanical behavior of the particle reinforced composite material. The proposed model is intuitive and straightforward capable of predicting composites' viscoelastic behavior in the time domain. The isotropization technique for non-uniform stress-strain fields and incremental Mori-Tanaka schemes for high volume fraction are adopted in this study. Effects of the imperfectly bonded interphase layer on the viscoelastic behavior on the dynamic mechanical behavior are also investigated. The proposed model is verified by the direct numerical simulation and DMA (dynamic mechanical analysis) experimental results. The proposed model is useful for multiscale analysis of viscoelastic composite materials, and it can also be extended to predict the nonlinear viscoelastic response of composite materials.

• ETRI Journal
• /
• 제46권1호
• /
• pp.59-70
• /
• 2024

We introduce a high-performance named entity recognition (NER) model for written and spoken language. To overcome challenges related to labeled data scarcity and domain shifts, we use transfer learning to leverage our previously developed KorBERT as the base model. We also adopt a meta-pseudo-label method using a teacher/student framework with labeled and unlabeled data. Our model presents two modifications. First, the student model is updated with an average loss from both human- and pseudo-labeled data. Second, the influence of noisy pseudo-labeled data is mitigated by considering feedback scores and updating the teacher model only when below a threshold (0.0005). We achieve the target NER performance in the spoken language domain and improve that in the written language domain by proposing a straightforward rollback method that reverts to the best model based on scarce human-labeled data. Further improvement is achieved by adjusting the label vector weights in the named entity dictionary.