• 디지털콘텐츠학회 논문지
• /
• 제19권8호
• /
• pp.1489-1497
• /
• 2018

온라인 뉴스에 개인의 참여가 활성화되면서 댓글의 중요성은 더욱 커지고 있다. 특히 이용자들에게 많은 공감을 받는 '베스트 댓글'은 주된 여론으로 인식되고 큰 영향력을 가진다. 따라서 본 연구는 온라인 뉴스 댓글 데이터를 이용하여 베스트 댓글의 특성을 알아보고자 하였다. 이를 위해 일반 댓글과 차이점을 보일 가능성이 있는 요소를 설정 후, 데이터를 수치화하여 일반 댓글과 베스트 댓글의 차이를 분석하였다. 본 연구는 최근 댓글 조작 등의 문제 해결에 실마리를 제공하고 개인 및 학술단체, 정부기관 등을 주체로 하여금 기초 자료로 활용될 것으로 기대된다.

• 한국콘텐츠학회논문지
• /
• 제20권5호
• /
• pp.513-524
• /
• 2020

최근 유튜브의 성장이 주목받고 있다. 유튜브는 콘텐츠를 소비하는 채널일 뿐만 아니라 소비자의 의사를 표시하는 공간이 되고 있다. 소비자들은 댓글이라는 형식으로 유튜브 안에서 그들의 의견을 표출한다. 이 연구는 유튜브에 업로드되어 시청되는 글로벌 기업가 연설문과 영상을 시청하고 작성한 댓글의 텍스트를 연구대상으로 한다. 텍스트 마이닝 소프트웨어인 Leximancer를 사용하여 각각의 연설문과 댓글을 대상으로 내용 분석(content analysis)을 시행하였다. 각 기업가 연설문의 주제는 대응분석(correspondence analysis)을 사용하여 분석하였고 기업가 개인의 성향과 특성과 관련 있는 주제를 도출하였다. 댓글에서는 각 연설문의 내용과는 관계없이 공통적으로 money, work, need의 주제를 발견하였다. 텍스트의 길이가 다름을 고려하여 추가로 중요도 지표 (Prominence Index) 분석을 실행하였고 연설문 내용과 시청자 댓글의 공통적인 주제를 구성하는 핵심어로 time, future, better, best, change, life, business, need를 도출하였다. 유튜브 연설문의 시청자는 인생과 시간, 미래, 고객의 니즈, 긍정적 변화에 대한 메시지에 대하여 동일한 주제로 반응하는 것으로 나타났다.

• 서비스연구
• /
• 제11권3호
• /
• pp.1-16
• /
• 2021

본 연구는 최근 많은 관심을 받고있는 크라우드펀딩을 더많은 창업 및 기업가가 활용하기 위해서는 크라우드펀딩의 성공을 가져온 요인을 이해하는 것이 중요하나, 동시에 정보의 비대칭성을 이유로 비판받고 있는바 펀드사와 자금 조달자를 위하여 비대칭 정보 소스에서 수집된 정보가 크라우드펀딩 성공에 미치는 영향을 제시하고자 한다. 또한, 금번 연구를 통하여 서비스 사이언스가 추구하는 융합의 사회적 장을 마련하고 실질적인 연구와 실천을 수행하며 제품의 서비스화, 서비스의 제품화를 주도하는 핵심 연구에 과학적 접근법이 필요한 상황에 대하여 지식기반서비스의 실제 사례를 도출하고자 하였다. 첫째, 크라우드펀딩 플랫폼 내에서 자금 제공자와 기금 모금자 사이의 상호작용의 중요성에 대하여 조사하였으며, "comments" 섹션에서는 자금 제공자와 기금 모금자 간의 의사소통을 통하여 비대칭 정보의 정도를 줄인다. 그리고, 잠정적인 자금 제공자의 측면에서는 투자에 대한 수익을 상실하지 않고 보다 확실한 정보를 수집하기 위하여 "comments"섹션을 최대한 활용해야 함을 알 수 있다. 둘째,"story"(비대칭소스)와 "comments" 섹션에서 감성을 기반으로 낙관적이고 비관적인 정보를 분석 하였다. 결론적으로 "comments" 섹션의 낙관적 정보와 비관적 정보가 크라우드펀딩 성공에 큰 영향을 미친다는 것을 보여주며, 상대적으로 "story" 섹션의 영향은 특정 조건에서의 비관적인 정보에만 한정됨을 알수 있다. 크라우드펀딩 플랫폼은 창업(startup) 및 기업가들이 비즈니스를 시작할 수 있는 혁신적이고 생산적인 방법으로, 정보가 크라우드펀딩의 성공에 중요한 역할을 하므로 최상의 의사소통 방법을 구축하는 데 집중하는 것이 필수적임을 확인할수 있다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권2호
• /
• pp.263-283
• /
• 2024

Users' comments after online shopping are critical to product reputation and business improvement. These comments, sometimes known as e-commerce reviews, influence other customers' purchasing decisions. To confront large amounts of e-commerce reviews, automatic analysis based on machine learning and deep learning draws more and more attention. A core task therein is sentiment analysis. However, the e-commerce reviews exhibit the following characteristics: (1) inconsistency between comment content and the star rating; (2) a large number of unlabeled data, i.e., comments without a star rating, and (3) the data imbalance caused by the sparse negative comments. This paper employs Bidirectional Encoder Representation from Transformers (BERT), one of the best natural language processing models, as the base model. According to the above data characteristics, we propose the F_MixBERT framework, to more effectively use inconsistently low-quality and unlabeled data and resolve the problem of data imbalance. In the framework, the proposed MixBERT incorporates the MixMatch approach into BERT's high-dimensional vectors to train the unlabeled and low-quality data with generated pseudo labels. Meanwhile, data imbalance is resolved by Focal loss, which penalizes the contribution of large-scale data and easily-identifiable data to total loss. Comparative experiments demonstrate that the proposed framework outperforms BERT and MixBERT for sentiment analysis of e-commerce comments.

• Journal of Information Processing Systems
• /
• 제19권5호
• /
• pp.641-651
• /
• 2023

Malicious hate speech and gender bias comments are common in online communities, causing social problems in our society. Gender bias and hate speech detection has been investigated. However, it is difficult because there are diverse ways to express them in words. To solve this problem, we attempted to detect malicious comments in a Korean hate speech dataset constructed in 2020. We explored bidirectional encoder representations from transformers (BERT)-based deep learning models utilizing hyperparameter tuning, data sampling, and logits ensembles with a label distribution. We evaluated our model in Kaggle competitions for gender bias, general bias, and hate speech detection. For gender bias detection, an F1-score of 0.7711 was achieved using an ensemble of the Soongsil-BERT and KcELECTRA models. The general bias task included the gender bias task, and the ensemble model achieved the best F1-score of 0.7166.

• Journal of Chest Surgery
• /
• 제45권4호
• /
• pp.213-224
• /
• 2012

Tetralogy of Fallot (TOF) is an index lesion for all paediatric and congenital heart surgeons. In designing an appropriate operation for children with TOF, the predicted postoperative physiology must be taken into account, both for the short and long term. A favourable balance between pulmonary stenosis (PS) and pulmonary insufficiency (PI) may be critical for preservation of biventricular function. A unified repair strategy to limit both residual PS and PI is presented, along with supportive experimental evidence. A strategy for dealing with coronary anomalies and some comments regarding best timing of operation are also included.

• 한국IT서비스학회지
• /
• 제14권2호
• /
• pp.177-193
• /
• 2015

A great number of customers, who want to watch movies usually check out online reviews before choosing what to watch a movie. The most representative online media that customers consult are portal sites and SNS (Social Network Service). Although there have been numerous studies on online eWOM (e-Word of Mouth) and the effects of online media in businesses, it remains a question that which media is best for WOM (Word of Mouth) when selecting movies. This research examines customer's intention for consulting eWOM and for watching movies according to the number and tendency of online replies. We have compared portal sites and SNS about information of movie. The study shows that a large number of positive replies can affect the intention for WOM and choosing movies. Facebook has more influence than portal sites when choosing what to watch when replies consist of large and positive comments. However, there is no difference between the two types of media when they consist of negative comments.

• 한국조경학회지
• /
• 제23권2호
• /
• pp.71-92
• /
• 1995

This study Is about the situation of contemporary landscape architecture design in Korea. The conceptual framework of interpretation for landscape architectural works is formulated to define the design trends of today′s professional works The study′s analytical method is composed of the survey of attitudes questionnaire, the analysis of design composition techniques and conclusions drawn from the comments of contemporary landscape architectural critics. Among 975 members of the Korean Institute of Landscape Architecture, 318 professionals responded to the survey questionnaire. The survey revealed that landscape architects in Korea are concerned about the profession′s reduced contributions to the sharping of the built environment. However, they are optimistic about its future. Ten top design works were selected from the surveys′ answers and were analyzed for the design characteristics which could best represent the trends of the last thirty years. Several design critiques and comments about the professional works were cited to reinforce the outlined "-ty" of the contemporary design phenomena. The study identified the following distinguished design trends of contemporary Korean landscape architectural works They are as follows: Themelessty, Powerity, Traditionality, Monumentality, Non-contextuality and Over-designity. These trends identify the overall prevailing conditions of Korean landscape architectural design over the past thirty years.

• 한국식생활문화학회지
• /
• 제34권2호
• /
• pp.184-200
• /
• 2019

This study aims to establish a strategy for selling kimchi to satisfy the needs and wants of European consumers through survey, FGI, and IDI. All survey participants were Europeans and were selected on the basis of their kimchi awareness. For the quantitative survey, 246 participants 15 years or older were selected; for FGI, 33 participants aged between 20 and 49 were selected; and for IDI, eight chefs were selected. The survey included four steps of 'Awareness-First try-Needs-Future consumption'. The participants got to know kimchi through the Internet; for example, by visiting Korea and Hallyu, and suggested positive comments, including 'fancy' and 'exotic.' While satisfaction with kimchi sharply decreased due to taste and appearance in the first try of kimchi, it is important to provide the best first-try experience. It also showed that important needs for kimchi are 'hot sensation', 'crunchiness' and 'presentation'. They gave a negative answer to combination with traditional cuisines of their own country, but provided positive comments for benchmarking the use of traditional pickled foods. This study suggests that it may be necessary to implement a two-track strategy by the traditional kimchi as a 'strategic product' and localized kimchi as a 'key product'.

• The Journal of Asian Finance, Economics and Business
• /
• 제9권10호
• /
• pp.189-200
• /
• 2022

The gig economy offers many advantages, such as flexibility, variety, independence, and lower cost. However, there are also safety concerns, lack of regulations, uncertainty, and unsatisfactory services, causing people to voice their opinion on social media. This paper aims to explore the sentiments of consumers concerning gig economy services (Grab, Foodpanda and Airbnb) through the analysis of social media. First, Vader Lexicon was used to classify the comments into positive, negative, and neutral sentiments. Then, the comments were further classified into three machine learning algorithms: Support Vector Machine, Light Gradient Boosted Machine, and Logistic Regression. Results suggested that gig economy services in Malaysia received more positive sentiments (52%) than negative sentiments (19%) and neutral sentiments (29%). Based on the three algorithms used in this research, LGBM has been the best model with the highest accuracy of 85%, while SVM has 84% and LR 82%. The results of this study proved the power of text mining and sentiment analysis in extracting business value and providing insight to businesses. Additionally, it aids gig managers and service providers in understanding clients' sentiments about their goods and services and making necessary adjustments to optimize satisfaction.