• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.432-436
• /
• 2023

초거대언어모델(LLM)의 도래에 따라 다양한 과업들이 도메인 관계 없이 제로샷으로 추론이 가능해짐에 따라서 LLM이 다양한 산업분야에 적용되고 있다. 대표적으로 ChatGPT와 GPT-4는 상용 API로 서비스를 제공하여 용이한 서비스 접근으로 다양한 이용층을 끌어들이고 있다. 그러나 현재 상용 API로 제공되고 있는 ChatGPT 및 GPT-4는 사용자의 대화 내역 데이터를 수집해 기업의 보안 문제를 야기할 수 있고 또한 생성된 결과물의 환각 문제로 인한 기업 문서의 신뢰성 저하를 초래할 수 있다. 특히 LLM 생성 글은 인간의 글과 유사한 수준으로 유창성을 확보한만큼 산업현장에서 LLM 작성 글이 판별되지 못할 경우 기업 활동에 큰 제약을 줄 수 있다. 그러나 현재 한국어 LLM 작성 글 탐지 서비스가 전무한 실정이다. 본 논문에서는 한국어 초거대언어모델 작성 글 판별기: KoCheckGPT 를 제안한다.KoCheckGPT는 산업현장에서 자주 사용되는 문어체, 개조식 글쓰기로 작성된 문서 도메인을 목표로 하여 글 전체와 문장 단위의 판별 정보를 결합하여 주어진 문서의 LLM 작성 여부를 효과적으로 판별한다. 다국어 LLM 작성 글 판별기 ZeroGPT와의 비교 실험 결과 KoCheckGPT는 우수한 한국어 LLM 작성 글 탐지 성능을 보였다.

• Journal of rehabilitation welfare engineering & assistive technology
• /
• v.10 no.3
• /
• pp.237-242
• /
• 2016

The major advantage of slip-ring technology in Spiral CT is that it facilitates continuous rotation of the x-ray tube, so that volume data can be acquired from a patient quickly. Not only for such a fast scan, but also for the dose reduction purpose, high signal-to-noise ratio and fast data acquisition system is required. In this study, we have built a multi-channel photodetector and multi-channel data acquisition system for CT application. The detector module consisted of CdWO4 crystal and Si photodiode in 16 channels. For the performance test of the preamplifier stage, both the transimpedance and switched integrator types are optimized for the photodetector modules. Switched integrator showed better noise performance in the limited bandwidth which is suitable for the current CT application. The control sequence for data acquisition and 20 bit ADC is designed with VHDL(Very High Speed Integrated Circuit Hardware Description Language) and implemented on FPGA(Field Programmable Gate Array) chip. Our Si photodiode detector module coupled to CdWO4 crystal showed comparable signal with other commercially available photodiode for CT. Switched integrator type showed higher SNR but narrower bandwidth compared to transimpedance preamplifier. Digital hardware is designed by FPGA, so that the control signal could be redesigned without hardware alteration.

• International Journal of Contents
• /
• v.13 no.1
• /
• pp.68-75
• /
• 2017

Interactive Typography is a process where a user communicates by interacting with text and a moving factor. This research covers interactive typography using real-time response to a user's gesture. In order to form a language-independent system, preprocessing of entered text data presents image data. This preprocessing is followed by recognizing the image data and the setting interaction points. This is done using computer vision technology such as the Harris corner detector and contour detection. User interaction is achieved using skeleton information tracked by a depth camera. By synchronizing the user's skeleton information acquired by Kinect (a depth camera,) and the typography components (interaction points), all user gestures are linked with the typography in real time. An experiment was conducted, in both English and Korean, where users showed an 81% satisfaction level using an interactive typography system where text components showed discrete movements in accordance with the users' gestures. Through this experiment, it was possible to ascertain that sensibility varied depending on the size and the speed of the text and interactive alteration. The results show that interactive typography can potentially be an accurate communication tool, and not merely a uniform text transmission system.

• Proceedings of the KOSOMBE Conference
• /
• v.1990 no.11
• /
• pp.11-14
• /
• 1990

The detection of pulmonary blood vessels is very difficult owing to their complex tree structures and different widths. In this paper, We propose a new detection algorithm. The motivation of this algorithm is that Han is the best detector. So, this algorithm is developed to imitate the human searching process. To realize it, the algorithm consist of two components. One is Pyramid Images whose one pixel is median value of four pixels of the previous low level. Searching gradually from high level to low level, We concentrate on global and main information of structure at the first. Then based on it, We search the detailed data in low level. The other is fuzzy logic which makes it easy to convert searching process expressed as human language into numeric multi_value. This algorithm showes speedy and robust results. But the more study on both human searching process and the detection of small part is needed.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.15 no.2
• /
• pp.263-268
• /
• 1997

Software structure must be dynamic to support new geospatial information sources and their object-oriented implementation on the Web. JAVA is interactive, platform idependent and object-oriented language and meets all needs on Internet GSIS. This paper introduce JAVA based GIS program to manipulates various geographic data on Internet, communicating interactively and transfer real-time data between server and client. This and this program analysis roof detector in all part of Pusan area and indicates the traffic states, road surface conditions, weather information, shortest cut, and road names in JAVA client windows. Also this study shows various techniques in expression real time traffic informations.

• Journal of Advanced Navigation Technology
• /
• v.15 no.2
• /
• pp.213-220
• /
• 2011

In this paper, the channel preprocessor with an area-efficient architecture is proposed for the MIMO symbol detector which can support four transmit and receive antennas. The proposed channel preprocessor can shrink the channel dimension to reduce the hardware complexity of the MIMO symbol detector. Also, the proposed channel preprocessor is implemented with very low complexity by using QR decomposition (QRD) and log-number system (LNS). By applying QRD and LNS to the nulling matrix calculation block, the numbers of matrix-multiplications and matrix-divisions are decreased and thus the complexity of the proposed channel preprocessor is significantly reduced. The proposed channel preprocessor was designed in a hardware description language (HDL) and synthesized to gate-level circuits using 0.13um CMOS standard cell library. With the proposed channel preprocessor, the number of logic gates for channel preprocessor is reduced by 20.2% compared with the conventional architecture.

• Journal of Broadcast Engineering
• /
• v.24 no.3
• /
• pp.495-505
• /
• 2019

Typical algorithms of deep learning include CNN(Convolutional Neural Networks), which are mainly used for image recognition, and RNN(Recurrent Neural Networks), which are used mainly for speech recognition and natural language processing. Among them, CNN is able to learn from filters that generate feature maps with algorithms that automatically learn features from data, making it mainstream with excellent performance in image recognition. Since then, various algorithms such as R-CNN and others have appeared in object detection to improve performance of CNN, and algorithms such as YOLO(You Only Look Once) and SSD(Single Shot Multi-box Detector) have been proposed recently. However, since these deep learning-based detection algorithms determine the success of the detection in the still images, stable object tracking and detection in the video requires separate tracking capabilities. Therefore, this paper proposes a method of combining Kalman filters into deep learning-based detection networks for improved object tracking and detection performance in the video. The detection network used YOLO v2, which is capable of real-time processing, and the proposed method resulted in 7.7% IoU performance improvement over the existing YOLO v2 network and 20 fps processing speed in FHD images.

• The Transactions of the Korea Information Processing Society
• /
• v.13 no.3
• /
• pp.148-155
• /
• 2024

Despite the increasing importance of irony and sarcasm detection in the field of natural language processing, research on the Korean language is relatively scarce compared to other languages. This study aims to experiment with various models for irony detection in Korean text. The study conducted irony detection experiments using KoBERT, a BERT-based model, and ChatGPT. For KoBERT, two methods of additional training on sentiment data were applied (Transfer Learning and MultiTask Learning). Additionally, for ChatGPT, the Few-Shot Learning technique was applied by increasing the number of example sentences entered as prompts. The results of the experiments showed that the Transfer Learning and MultiTask Learning models, which were trained with additional sentiment data, outperformed the baseline model without additional sentiment data. On the other hand, ChatGPT exhibited significantly lower performance compared to KoBERT, and increasing the number of example sentences did not lead to a noticeable improvement in performance. In conclusion, this study suggests that a model based on KoBERT is more suitable for irony detection than ChatGPT, and it highlights the potential contribution of additional training on sentiment data to improve irony detection performance.

• The KIPS Transactions:PartA
• /
• v.11A no.1
• /
• pp.81-88
• /
• 2004

In this paper, we designed the outer encoder/decoder for the terrestrial DMB that is an advanced digital broadcasting standard, implemented, and verified by using ALTERA FPGA. In the encoder part, it was created the parity bytes (16 bytes) from the input packet (188by1e) of MPEG-2 TS and the encoded data was distributed output by the convolutional interleaver for Preventing burst errors. In the decoder part, It was proposed the algorithm that detects synchronous character suitable to DMB in transmitted data from the encoder. The circuit complexity in RS decoder was reduced by applying a modified Euclid's algorithm. This system has a capability to correct error of the maximum 8 bytes in a packet. After the outer encoder/decoder algorithm was verified by using C language, described in VHDL and implemented in the ALTERA FPGA chips.

• Journal of Radiation Protection and Research
• /
• v.22 no.3
• /
• pp.143-151
• /
• 1997

As the pulse heights from a NaI(Tl) detector vary with the temperature of the measuring environment a significant change in temperature may affect the energy calibration of the spectrometer. The auto-adjustment of the channel corresponding to a pulse heights can be achieved by introducing an external reference source to compensate the temperature dependency of pulse heights, but unfavorable increases of the Compton continuum are caused due to the external source. In this study, the total absorption peaks dominant in the typical environmental gamma spectrum-239 keV from $^{212}Pb$, 351 keV from $^{214}Pb$, 1460 keV from $^{40}K$ and 2614 keV from $^{208}Tl$ for examples - were used as reference in the correction of energy calibration. With these peaks, the program to calibrate the energy of the s spectrum was developed using Microsoft Visual Basic language. The program developed here was applied to the environmental spectra measured at intervals of 30 minutes in the temperature range of from $-20^{\circ}C$ to $10^{\circ}C$ to demonstrate the validity and applicability. As a result of the test, the correction scheme appeared to be effective in the temperature changes encountered in the usual environment.