• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.3
• /
• pp.125-132
• /
• 2022

Sentence compression is a natural language processing task that generates concise sentences that preserves the important meaning of the original sentence. For grammatically appropriate sentence compression, early studies utilized human-defined linguistic rules. Furthermore, while the sequence-to-sequence models perform well on various natural language processing tasks, such as machine translation, there have been studies that utilize it for sentence compression. However, for the linguistic rule-based studies, all rules have to be defined by human, and for the sequence-to-sequence model based studies require a large amount of parallel data for model training. In order to address these challenges, Deleter, a sentence compression model that leverages a pre-trained language model BERT, is proposed. Because the Deleter utilizes perplexity based score computed over BERT to compress sentences, any linguistic rules and parallel dataset is not required for sentence compression. However, because Deleter compresses sentences only considering perplexity, it does not compress sentences by reflecting the linguistic information of the words in the sentences. Furthermore, since the dataset used for pre-learning BERT are far from compressed sentences, there is a problem that this can lad to incorrect sentence compression. In order to address these problems, this paper proposes a method to quantify the importance of linguistic information and reflect it in perplexity-based sentence scoring. Furthermore, by fine-tuning BERT with a corpus of news articles that often contain proper nouns and often omit the unnecessary modifiers, we allow BERT to measure the perplexity appropriate for sentence compression. The evaluations on the English and Korean dataset confirm that the sentence compression performance of sentence-scoring based models can be improved by utilizing the proposed method.

• Korea Science and Art Forum
• /
• v.35
• /
• pp.281-292
• /
• 2018

Crowd funding is also called social funding because of SNS that it helps early start-up founder and makers to raise money for idea product production. Recently, the funding platform has recorded high growth rates. As a result, the government in Korea has introduced various support policies for the crowd funding. The purpose of this study is to develop a diagnostic design guideline for product design oriented makers based on the historical situation. The paper writer applied literature survey and expert interview as research methods. The literature survey focused on internet news and previous research studies. The expert interview was conducted for 10 specialist people and divided for the second time. As a result of the text survey, the current guideline was lacking in design and in detail. Researchers have been informed through previous paper that information transfer text and images are important factors for funding success. In the first interview with seven special participants, we made a draft design guideline for social funding with a two-step process and nine themes. We, research and three professional people having a evaluation experience, conducted verification and supplementation for establishing the design guider with a three-step process and eight themes in the next interview. The design guideline for crowd funding, it can be used by money funding manager apart from design makers. Through the results of this paper, researchers are expected to prevent problems and contribute to healthy crowd funding ecosystem development.

• Journal of Internet Computing and Services
• /
• v.25 no.1
• /
• pp.57-67
• /
• 2024

This research investigates training methods for large language models to accurately identify sentiments and comprehend information about increasing and decreasing fluctuations in the financial domain. The main goal is to identify suitable datasets that enable these models to effectively understand expressions related to financial increases and decreases. For this purpose, we selected sentences from Wall Street Journal that included relevant financial terms and sentences generated by GPT-3.5-turbo-1106 for post-training. We assessed the impact of these datasets on language model performance using Financial PhraseBank, a benchmark dataset for financial sentiment analysis. Our findings demonstrate that post-training FinBERT, a model specialized in finance, outperformed the similarly post-trained BERT, a general domain model. Moreover, post-training with actual financial news proved to be more effective than using generated sentences, though in scenarios requiring higher generalization, models trained on generated sentences performed better. This suggests that aligning the model's domain with the domain of the area intended for improvement and choosing the right dataset are crucial for enhancing a language model's understanding and sentiment prediction accuracy. These results offer a methodology for optimizing language model performance in financial sentiment analysis tasks and suggest future research directions for more nuanced language understanding and sentiment analysis in finance. This research provides valuable insights not only for the financial sector but also for language model training across various domains.

• Information Systems Review
• /
• v.20 no.2
• /
• pp.139-161
• /
• 2018

As Korea became one of the oldest countries in the world, successful aging emerged as an important issue to individuals as well as to society. This study aims to determine not only the Korean seniors' happiness and unhappiness factors but also the means to enhance their happiness and deal with unhappiness. We collected news articles related to the happiness and unhappiness of seniors with nine keywords based on Alderfer's ERG Theory. We then applied a topic modeling technique, Latent Dirichlet Allocation, to examine the main issues underlying the seniors' happiness and unhappiness. According to the analysis, we investigated the conditions of happiness and unhappiness by inspecting the topics based on each keyword. We also conducted a detailed analysis based on the main factors from topic modeling. We proposed specific ways to increase and overcome the happiness and unhappiness of seniors, respectively, in terms of government, corporate, family, and other social welfare organizations. This study indicates the major factors that affect the happiness and unhappiness of seniors. Specific methods to boost happiness and relief unhappiness are suggested from the additional analysis.

• Journal of Navigation and Port Research
• /
• v.48 no.1
• /
• pp.34-48
• /
• 2024

As global supply chain risks proliferate unpredictably, the high interdependence of port and logistics industry intensifies the risk burden. This study conducted fundamental research to explore diverse safety issues in domestic ports. Utilizing news article data about Busan Port, we employed LDA topic modeling and time-series linear regression to understand key safety trends. Over the past 30 years, Busan Port faced nine major safety issues-maritime safety, import cargo inspection, labor strikes, and natural disasters emerged cyclically. Major port safety issues in Busan Port are primarily characterized by an unpredictable nature, falling under socio-environmental and natural phenomena types, indicating a significant impact of global uncertainty. Therefore, systematic policies need to be formulated based on identified port safety issues to enhance port safety in Busan Port. Additionally, there is a need to strengthen the resilience of port safety for unpredictable risk situations. In conclusion, advanced research activities are necessary to promote port safety enhancement in response to dynamically changing social conditions.

• The Journal of Bigdata
• /
• v.9 no.1
• /
• pp.75-88
• /
• 2024

Recently, there have been cases reported in the news of individuals experiencing symptoms of food poisoning after consuming raw beef purchased from online platforms, or reviews claiming that cherry tomatoes tasted bitter. This suggests the potential for analyzing food reviews on online platforms to detect food hazards, enabling government agencies, food manufacturers, and distributors to manage consumer food safety risks. This study proposes a classification model that uses sentiment analysis and large language models to analyze food reviews and detect negative ones, multi-labeling key food safety hazards (food poisoning, spoilage, chemical odors, foreign objects). The sentiment analysis model effectively minimized the misclassification of negative reviews with a low False Positive rate using a 'funnel' model. The multi-labeling model for food safety hazards showed high performance with both recall and accuracy over 96% when using GPT-4 Turbo compared to GPT-3.5. Government agencies, food manufacturers, and distributors can use the proposed model to monitor consumer reviews in real-time, detect potential food safety issues early, and manage risks. Such a system can protect corporate brand reputation, enhance consumer protection, and ultimately improve consumer health and safety.

• Journal of Korean Theatre Studies Association
• /
• no.49
• /
• pp.5-47
• /
• 2013

Among the styles of performing arts, perhaps the genre that has attracted the largest audience would be musical. Popularity of musical has brought diverse changes in our performing arts market, and, upon emerging another musical genre, called 'Korean Traditional Musical Comedy,' it has been well-received by the audiences. 'Korean Traditional Musical Comedy' is a word that are formed by merging two other terms such as 'Korean Traditional Music' and 'Musical (Comedy).' In the meantime, however, it has yet some problems in order to be defined as the genre that has concrete concepts. It is because the term such as Korean Traditional Musical Comedy was created being closely associated with a marketing purpose rather than a term that defines the characteristics of a genre of performing arts. Although this new musical genre has drawn attentions of many audiences by adding 'Musical Comedy' to 'Korean Traditional Music' that was not quite popular to the public, it still does not have any established forms so that there is a fine line between "Korean Traditional Musical Comedy" and another genre like traditional style folk opera ("Changgeuk"). Looking at the characteristics of the musical work called 'Korean Traditional Musical Comedy, in general, first of all, it is a performance where music and drama are played. Here, the distinctive characteristic of this musical is that 'Korean Traditional Music' is sung. And the kinds of Korean traditional musics being sung are mainly Pansori (dramatic story-singing) and folk-songs, and, in most cases, Korean traditional musical instruments are being used as accompanying music. In this paper, the researcher investigated the aspects of experiment centering around Korean Musical Group, Taroo. These days, various experiments has been repeated not only for the works of Taroo but other musical work presently called 'Korean Traditional Musical Comedy' also. Having encompassed overall performance factors including use of musical instruments, dance, acting, materials for drama as well as music in drama, the researcher has gone through experiments repeatedly. Meanwhile, however, the subject matters that make 'Korean Traditional Musical Comedy' mostly attractive to the audiences are music and songs or sounds. ["Sori" also called "Chang" (唱)] Particularly, under the current situation of our musicals, the role of "Sori" is extremely important. The factor that plays absolutely most important role in acceptance and transformation of "Sori" is the created Pansori. Since the created Pansori is composed with new rhythmic patterns and new narrative poems, it tells the present story. Also it draws good responses from the audiences owing to easy conveyance of dialogues. And, its new style brings diversification to organization of musical instruments, so then this leads to the arrangements of music for Korean traditional music instruments, as well as instrumental music ensemble, orchestra, and jazz band, etc. Likewise, upon appearing creative musics in 'Korean Traditional Musical Comedy,' professional music and vocal compositions have begun to emerge naturally. And, the song specialist and writer, of course, staffs including direction, lighting, and sounds, etc are required. That is, professional composition method are forced to be introduced to all areas. Other than this, there are many music pieces which are based on our unique songs and sounds ("Sori") and such traditional factors as use of lead singer for ceremony or chorus, and the method that puts weight on Pansori. Accordingly many things accomplished. However, it is required that 'Korean Traditional Musical Comedy' go through numerous discussions and more experiments. Above all, the most important things are the role of actor and actress, and their changes, and training of actor and actress further. Good news is there are good audience responses. 'Korean Traditional Musical Comedy' is an open genre. As musicals are divided into several domains according to the characteristics thereof, 'Korean Traditional Musical Comedy' will be able to show its distinctive features in various styles according to embodiment.

• The Journal of Society for e-Business Studies
• /
• v.25 no.3
• /
• pp.41-61
• /
• 2020

China's rapid development and commercialization of high-tech technologies in the fourth industrial revolution has led to effective technology exchanges between Korean and Chinese firms becoming more important to Korea's mid-term and long-term industrial development. However, there is still a lack of empirical research on how technology exchanges between Korean and Chinese firms proceed and their effectiveness. In response, this study conducted a social network analysis based on text mining data of Korea-China business technology exchange and cooperation articles introduced in the news from 2018 to March 2020 on the current status and effects of Korea-China technology exchanges related to the fourth industrial revolution, and conducted a regression analysis how network centrality effect on the firm performance. According to the results, most of the Korean major electronic firms are actively networking with Chinese firms and institutions, showing high centrality in the centrality index. Korean telecommunication firms showed high betweenness centrality and subgraph centrality, and Korean Internet service providers and broadcasting contents firms showed high eigenvector centrality. In addition, Chinese firms showed higher betweenness centrality than Korean firms, and Chinese service firms showed higher closeness centrality than manufacturing firms. As a result of regression analysis, this network centrality had a positive effect on firm performance. To the best of our knowledge, this is the first to analyze the impact of the technical cooperation between Korean and Chinese firms under the fourth industrial revolution context. This study has theoretical implications that suggested the direction of social network analysis-based empirical research in global firm cooperation. Also, this study has practical implications that the guidelines for network analysis in setting the direction of technical cooperation between Korea and China by firms or governments.

• Journal of The Korean Association For Science Education
• /
• v.32 no.4
• /
• pp.775-788
• /
• 2012

This study was done at a scientific camp sponsored by Nara Women's University Secondary School, Japan. In this school, $10^{th}$ grade students from 4 East Asian countries: Korea, Japan, Taiwan, and Singapore, participated. We made a research on students' perceptions about nuclear energy. Sample populations include 77 students in total, with 12 Korean, 46 Japanese, 9 Taiwanese and 10 Singaporean students. Overall perceptions comparison about nuclear energy shows average values from the order of highest Korea, Taiwan, Singapore, and to lowest, Japan. We implemented a T-test to identify perception differences about nuclear energy, with one group that include 3 countries (Korea, Taiwan and Singapore) and another group that includes all the Japanese students. T-test results of perceptions about nuclear energy shows students from the 3 countries of Korea, Taiwan and Singapore having higher average than Japanese students. (p<.05). Korean average scores regarding overall perceptions about nuclear energy show as the highest in all 4 East Asian countries and also highest in all subcategories. On the contrary in Japan, they have lower and negative perceptions of nuclear energy. In spite of these facts, perceptions of Japanese students about nuclear energy seem lowest and negative mainly because of the recent Fukushima nuclear power plant disaster, caused by the tsunami and its subsequent damages and fears of radiation leaks, etc. This shows that negative information about future disasters and its resulting damages like the Chernobyl nuclear accident could influence more on people's risk perception than general information like nuclear energy-related technologies or the news that the plant is operating normally, etc. Even if the possibility of this kind of accident is very low, just one accident could bring abnormal risks to technology itself. This strong signal makes negative image and strengthens its perceptions to the people. This could bring a stigma about nuclear energy. This study shows that Government's policy about the highest priority for nuclear energy safety is most important. As long as such perception and decision are fixed, we found that it might not be easy to get changed again because they were already fortified and maintained.

• Journal of Intelligence and Information Systems
• /
• v.28 no.2
• /
• pp.191-206
• /
• 2022

Recently, it is a de-facto approach to utilize a pre-trained language model(PLM) to achieve the state-of-the-art performance for various natural language tasks(called downstream tasks) such as sentiment analysis and question answering. However, similar to any other machine learning method, PLM tends to depend on the data distribution seen during the training phase and shows worse performance on the unseen (Out-of-Distribution) domain. Due to the aforementioned reason, there have been many efforts to develop domain-specified PLM for various fields such as medical and legal industries. In this paper, we discuss the training of a finance domain-specified PLM for the Korean language and its applications. Our finance domain-specified PLM, KB-BERT, is trained on a carefully curated financial corpus that includes domain-specific documents such as financial reports. We provide extensive performance evaluation results on three natural language tasks, topic classification, sentiment analysis, and question answering. Compared to the state-of-the-art Korean PLM models such as KoELECTRA and KLUE-RoBERTa, KB-BERT shows comparable performance on general datasets based on common corpora like Wikipedia and news articles. Moreover, KB-BERT outperforms compared models on finance domain datasets that require finance-specific knowledge to solve given problems.