• Journal of The Korean Association of Information Education
• /
• v.26 no.3
• /
• pp.197-207
• /
• 2022

In this paper, we develop a machine learning based block code generation and recommendation model for the purpose of reducing cognitive load of learners during coding education that learns the learner's block that has been made in the block programming environment using natural processing model and fine-tuning and then generates and recommends the selectable blocks for the next step. To develop the model, the training dataset was produced by pre-processing 50 block codes that were on the popular block programming language web site 'Entry'. Also, after dividing the pre-processed blocks into training dataset, verification dataset and test dataset, we developed a model that generates block codes based on LSTM, Seq2Seq, and GPT-2 model. In the results of the performance evaluation of the developed model, GPT-2 showed a higher performance than the LSTM and Seq2Seq model in the BLEU and ROUGE scores which measure sentence similarity. The data results generated through the GPT-2 model, show that the performance was relatively similar in the BLEU and ROUGE scores except for the case where the number of blocks was 1 or 17.

• Journal of the Institute of Convergence Signal Processing
• /
• v.24 no.3
• /
• pp.160-165
• /
• 2023

In this study, we propose an SNS sentence writing assistance system that utilizes YOLO and GPT to assist users in writing texts with images, such as SNS. We utilize the YOLO model to extract objects from images inserted during writing, and also extract meta-information such as GPS information and creation time information, and use them as prompt values for GPT. To use the YOLO model, we trained it on form image data, and the mAP score of the model is about 0.25 on average. GPT was trained on 1,000 blog text data with the topic of 'restaurant reviews', and the model trained in this study was used to generate sentences with two types of keywords extracted from the images. A survey was conducted to evaluate the practicality of the generated sentences, and a closed-ended survey was conducted to clearly analyze the survey results. There were three evaluation items for the questionnaire by providing the inserted image and keyword sentences. The results showed that the keywords in the images generated meaningful sentences. Through this study, we found that the accuracy of image-based sentence generation depends on the relationship between image keywords and GPT learning contents.

• Annual Conference on Human and Language Technology
• /
• 2020.10a
• /
• pp.495-500
• /
• 2020

뉴스 기사는 반드시 객관적이고 넓은 시각으로 정보를 전달하지 않는다. 따라서 뉴스 기사를 기존의 추천 시스템과 같이 개인의 관심사나 사적 정보를 바탕으로 선별적으로 추천하는 것은 바람직하지 않다. 본 논문에서는 최대한 객관적으로 다양한 시각에서 비슷한 사건과 인물에 대해서 판단할 수 있도록 유사도 기반의 기사 추천 모델을 제시한다. 길이가 긴 문서 사이의 유사도를 측정하기 위해 GPT2 [1]언어 모델을 활용했다. 이 과정에서 단방향 디코더 모델인 GPT2 [1]의 단점을 추가 학습으로 개선했으며, 저장 공간의 효율과 핵심 문단 추출을 위해 BM25 [2]함수를 사용했다. 그리고 준 지도 학습 [3]을 통해 유사도 레이블링이 되어있지 않은 최신 뉴스 기사에 대해서도 자가 학습을 진행했으며, 이와 함께 길이가 긴 문단에 대해서도 효과적으로 학습할 수 있도록 문장 길이를 기준으로 3개의 단계로 나누어진 커리큘럼 학습 [4]방식을 적용했다.

• Korean Journal of Soil Science and Fertilizer
• /
• v.3 no.1
• /
• pp.43-48
• /
• 1970

In order to investigate the inter-relation with the growth of the barley root and GOT and GPT activities the growth of root and GOT, GPT activities were measured those which have been supplied various nitrogen compounds ($NO_3-N$, $NH_4-N$, Urea, and Amino acid). The results obtained are summarized as follows: 1. Growth of barley root supplied with $NH_4-N$ is generally increased in length and weight compared with that of the root fertilized by $NH_4-N$. 2. The above-mentioned root with $NH_4-N$ is not only decreased in its weight and length but also is apt to inhibited its growth, in compared with the root provided with $NO_3-N$. 3. The activities of GOT and GPT for the root fertilized by $NH_4-N$, the badly grown root is generally increased, while of the root supplied with $NO_3-N$ is decreased compared with that of the root fertilized by $NH_4-N$. 4. The activities of GOT and GPT for the root provided with amino acid known as the considerable growth inhibiting compound for rice is generally decreased, while that of the badly known-grown root is increased. 5. The activities of GOT and GPT in the supernatant fraction of the barley is for the most part, high and low in the mitocondrial fraction.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.12
• /
• pp.53-59
• /
• 2021

In this paper, we propose a BERGPT-chatbot, a domestic AI chatbot that can alleviate negative emotions based on text input such as 'Replika'. We made BERGPT-chatbot into a chatbot capable of mitigating negative emotions by pipelined two models, KR-BERT and KoGPT2-chatbot. We applied a creative method of giving emotions to unrefined everyday datasets through KR-BERT, and learning additional datasets through KoGPT2-chatbot. The development background of BERGPT-chatbot is as follows. Currently, the number of people with depression is increasing all over the world. This phenomenon is emerging as a more serious problem due to COVID-19, which causes people to increase long-term indoor living or limit interpersonal relationships. Overseas artificial intelligence chatbots aimed at relieving negative emotions or taking care of mental health care, have increased in use due to the pandemic. In Korea, Psychological diagnosis chatbots similar to those of overseas cases are being operated. However, as the domestic chatbot is a system that outputs a button-based answer rather than a text input-based answer, when compared to overseas chatbots, domestic chatbots remain at a low level of diagnosing human psychology. Therefore, we proposed a chatbot that helps mitigating negative emotions through BERGPT-chatbot. Finally, we compared BERGPT-chatbot and KoGPT2-chatbot through 'Perplexity', an internal evaluation metric for evaluating language models, and showed the superity of BERGPT-chatbot.

• Journal of Medicine and Life Science
• /
• v.20 no.2
• /
• pp.53-59
• /
• 2023

This study explores the potential use of artificial intelligence (AI)-based services, specifically ChatGPT-3.5, in medical education. The application of this technology is acknowledged as a valuable tool for simulating authentic clinical scenarios and enhancing learners' diagnostic and communication skills. To construct a case, students received ChatGPT training using a clinical ethics casebook titled "Clinical Ethics Cases and Commentaries for Medical Students and Physicians." Subsequently, a role-play script was generated based on this training. The initial draft of the script was reviewed by two medical professors and was further optimized using ChatGPT-3.5. Consequently, a comprehensive role-play script, accurately reflecting real-world clinical situations, was successfully developed. This study demonstrates the potential for effectively integrating AI technology into medical education and provides a solution to overcome limitations in developing role-play scripts within conventional educational settings. However, the study acknowledges that AI cannot always generate flawless role-play scripts and recognizes the necessity of addressing these limitations and ethical concerns. The research explores both the potential and limitations of employing AI in the early stages of medical education, suggesting that future studies should focus on overcoming these limitations while further investigating the potential applications of AI in this field.

• Korean Journal of Community Nutrition
• /
• v.3 no.4
• /
• pp.574-582
• /
• 1998

This study was carried out to discover the effects of eating habits and health-related life style on blood pressure, $\gamma$-Glutamic acid Peptide Transferase($\gamma$-GPT), blood glucose and High Density Lipoprotein-Cholesterol(HDL-C). 185 subjects(85 male, 100 female) were selected, who were living in the Cheonju area aged 40#s to 60#s. The mean systollic blood pressure(SBP), diastollic blood pressure (DBP), $\gamma$-GPT, fasting blood sugar(FBS) and HDL-C for all the subjects were 118mmHg, 77mmHg, 281U/l, 90mg/dl and 45mg/dl, respectively. The SBP and DBP for subuects over 60 years old were 126mmHg and 81mmHg and were significantly higher than subjects in their 40#s and 50#s(p<0.001, p<0.005). The HDL-C of the group that rarely ate breakfast was 57mg/dl and that was significantly higher than the 44mg/dl scored by those who ate breakfast everyday(p<0.05). The SBP for subjects having a snack 2-3 times/week was 125mmHg and that was significantly higher than the 114mmHg of those having a snack everyday(p<0.05). The $\gamma$-GPT for subjects consuming alcohol everyday was 44IU/L and that was significantly higher than 18IU/I of the non-drinking group(p<0.001). The $\gamma$-GPT of light smokers was 53IU/I and that was significantly higher than the 22IU/I for non-smoking participants(p<0.001). The DBP, SBP, $\gamma$-GPT, FBS and HDL-C related to exercise not significantly different. The SBP(p<0.001) and DBP(p-0.01) between age group was positively correlated. The $\gamma$-GPT between drinking frequency(p<0.001), drinking quantity(p<0.05), and smoking(p<0.05) was also positively correlated. The FBS between exercises had a negative correlation(p<0.05), and the HDL-C between breakfast had a negative correlation(p<0.05). These results indicate that decreasing drinking and smoking, when combined with appropriate exercise, will decrease the $\gamma$-GPT and fasting blood sugar level, and help preventing adult diseases.

• Journal of Aquaculture
• /
• v.18 no.3
• /
• pp.196-199
• /
• 2005

Farmed red sea breams, Pagrus major, were fed for 60 days with pellets containing different concentrations of benzo(a)pyrene (0, 0.2, 2, 20 mg/kg) to generate a biomarker of the chemical toxicity in the fish. The fish exposed to the chemical concentrations did not show any significant difference in the weight gain, conditioning, factor, and hepatosomatic index. However, some haematological parameters, such as glucose, calcium, magnesium, GOT (glutamic oxalate transaminase), and GPT (glutamic pyruvate transaminase) were influenced by the chemical exposure. Of them, two enzymes, GOT and GPT, increased significantly 60 days after the exposure in a way of concentration dependence (P<0.05). In the study of ecotoxicological biomarker, sensitivity to adverse environments is one of the key available factors. The fish changes in GOT and GPT were an earlier and reliable sign of the fish response against the chemical exposure, rendering the two enzymatic factors as a useful biomarker at least to benzo(a)pyrene exposure in the farming waters.

• Journal of Information Science Theory and Practice
• /
• v.12 no.2
• /
• pp.79-95
• /
• 2024

The new AI based conversational chatbot, ChatGPT, launched in November 2022, is causing a stir. There are many opinions about this being a 'threat or a promise,' and thus it is important to understand what has been said about this tool and, based on the growing literature that has emerged on the subject, demystify its effective impact on society. To analyse this impact, a systematic literature review with the support of the preferred reporting items for systematic reviews and meta-analysis protocol was used. The data, scientific documents, were collected using the main scientific databases - SCOPUS and Web of Science - and the results were presented based on a bibliometric and thematic exploration of content. The main findings indicate that people are increasingly using this chatbot in more diverse areas. Therefore, this study contributes at the practical level, aiming to enlighten people in general - both in professional and personal life - about this tool and its impacts. Also, it contributes at the theoretical level, which involves expanding understanding and elucidation of the impacts of ChatGPT in different areas of study.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.07a
• /
• pp.541-544
• /
• 2023

LLM(Largescale Language Model)의 성능 향상을 위한 비용 효율적인 방법으로 ChatGPT, GPT-4와 같은 초거대 모델의 output에 대해 SLM(Small Language Model)을 finetune하는 방법이 주목받고 있다. 그러나, 이러한 접근법은 주로 범용적인 지시사항 모델을 위한 학습 방법으로 사용되며, 제한된 특정 도메인에서는 추가적인 성능 개선의 여지가 있다. 본 연구는 특정 도메인(Writing Assistant)에서의 성능 향상을 위한 새로운 방법인 Self-Guided Approach를 제안한다. Self-Guided Approach는 (1) LLM을 활용해 시드 데이터에 대해 도메인 특화된 metric(유용성, 관련성, 정확성, 세부사항의 수준별) 점수를 매기고, (2) 점수가 매겨진 데이터와 점수가 매겨지지 않은 데이터를 모두 활용하여 supervised 방식으로 SLM을 미세 조정한다. Vicuna에서 제안된 평가 방법인, GPT-4를 활용한 자동평가 프레임워크를 사용하여 Self-Guided Approach로 학습된 SLM의 성능을 평가하였다. 평가 결과 Self-Guided Approach가 Self-instruct, alpaca와 같이, 생성된 instruction 데이터에 튜닝하는 기존의 훈련 방법에 비해 성능이 향상됨을 확인했다. 다양한 스케일의 한국어 오픈 소스 LLM(Polyglot1.3B, PolyGlot3.8B, PolyGlot5.8B)에 대해서 Self-Guided Approach를 활용한 성능 개선을 확인했다. 평가는 GPT-4를 활용한 자동 평가를 진행했으며, Korean Novel Generation 도메인의 경우, 테스트 셋에서 4.547점에서 6.286점의 성능 향상이 발생했으며, Korean scenario Genration 도메인의 경우, 테스트 셋에서 4.038점에서 5.795 점의 성능 향상이 발생했으며, 다른 유사 도메인들에서도 비슷한 점수 향상을 확인했다. Self-Guided Approach의 활용을 통해 특정 도메인(Writing Assistant)에서의 SLM의 성능 개선 가능성을 확인했으며 이는 LLM에 비용부담을 크게 줄이면서도 제한된 도메인에서 성능을 유지하며, LLM을 활용한 응용 서비스에 있어 실질적인 도움을 제공할 수 있을 것으로 기대된다.