• Proceedings of the Korea Technology Innovation Society Conference
• /
• 2015.11a
• /
• pp.927-945
• /
• 2015

본 연구는 과학 커뮤니케이션 분야에서 널리 쓰이고 있는 뉴스 내용분석 방법론이 하이프 사이클 모델에 부합하는 지를 탐색적으로 살펴보고자 한다. 즉 과학기술 뉴스 내용분석이 하이프 사이클 모델에서 설명하는 사회적 가시성의 실체적 파악을 위한 기술기획의 유용한 보완적 방법론으로 쓰일 수 있음을 밝히는데 본 연구의 의의가 있다. 이를 위해 대표적인 유망기술로 클라우드 컴퓨팅을 대상으로 뉴스 내용분석을 수행하였다. 분석의 초점은 클라우드 컴퓨팅 기술 관련 뉴스의 빈도, 보도태도(긍정, 중립, 부정), 5가지 뉴스 프레임 관점에서 분석이 이루어졌고, 뉴스 보도경향이 하이프 사이클 흐름을 따라가는 지를 살펴보았다. 종합지 경제지와 IT전문지를 대상으로 한 뉴스 내용분석 결과는 뉴스 빈도, 보도 태도, 뉴스 프레임 모두 하이프 사이클의 흐름을 따르고 있었으며, 특히 2014년 이후의 흐름은 하이프 사이클 상에서 기대붕괴 지점을 지나 현실인식의 지점으로 진화되는 시점임을 추론할 수 있었다. 본 연구결과는 최근 확산되고 있는 텍스트 마이닝, 감성어 자동식별 분석 기술 등과 접목하여 사회적 맥락 파악을 위한 기술기획 분석의 보완적 방법론으로 기여할 수 있을 것으로 판단된다.

• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.320-325
• /
• 2021

BERT는 사전 학습 단계에서 다음 문장 예측 문제와 마스킹된 단어에 대한 예측 문제를 학습하여 여러 자연어 다운스트림 태스크에서 높은 성능을 보였다. 본 연구에서는 BERT의 사전 학습 문제 중 다음 문장 예측 문제에 대해 주목했다. 다음 문장 예측 문제는 자연어 추론 문제와 질의 응답 문제와 같이 임의의 두 문장 사이의 관계를 모델링하는 문제들에 성능 향상을 위해 사용되었다. 하지만 BERT의 다음 문장 예측 문제는 두 문장을 특수 토큰으로 분리하여 단일 문자열 형태로 모델에 입력으로 주어지는 cross-encoding 방식만을 학습하기 때문에 문장을 각각 인코딩하는 bi-encoding 방식의 다운스트림 태스크를 고려하지 않은 점에서 아쉬움이 있다. 본 논문에서는 기존 BERT의 다음 문장 예측 문제를 확장하여 bi-encoding 방식의 다음 문장 예측 문제를 추가적으로 사전 학습하여 단일 문장 분류 문제와 문장 임베딩을 활용하는 문제에서 성능을 향상 시키는 Bi-Cross 사전 학습 기법을 소개한다. Bi-Cross 학습 기법은 영화 리뷰 감성 분류 데이터 셋인 NSMC 데이터 셋에 대해 학습 데이터의 0.1%만 사용하는 학습 환경에서 Bi-Cross 사전 학습 기법 적용 전 모델 대비 5점 가량의 성능 향상이 있었다. 또한 KorSTS의 bi-encoding 방식의 문장 임베딩 성능 평가에서 Bi-Cross 사전 학습 기법 적용 전 모델 대비 1.5점의 성능 향상을 보였다.

• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.173-177
• /
• 2021

분절을 통한 양질의 입력 자질을 구성하는 것은 언어모델의 문장에 대한 이해도를 높이기 위한 필수적인 단계이다. 분절은 문장의 의미를 이해하는 데 있어 중요한 역할을 하기 때문이다. 따라서, 한국어 문장 분류 태스크를 수행함에 있어 한국어의 특징에 맞는 분절 기법을 선택하는 것은 필수적이다. 명확한 판단 기준 마련을 위해, 우리는 한국어 문장 분류 태스크에서 가장 효과적인 분절 기법이 무엇인지 감성 분석, 자연어 추론, 텍스트 간 의미적 유사성 판단 태스크를 통해 검증한다. 이 때 비교할 분절 기법의 유형 분류 기준은 언어학적 단위에 따라 어절, 형태소, 음절, 자모 네 가지로 설정하며, 분절 기법 외의 다른 실험 환경들은 동일하게 설정하여 분절 기법이 문장 분류 성능에 미치는 영향만을 측정하도록 한다. 실험 결과에 따르면 자모 단위의 분절 기법을 적용한 모델이 평균적으로 가장 높은 성능을 보여주며, 반복 실험 간 편차가 적어 일관적인 성능 결과를 기록함을 확인할 수 있다.

• Journal of Korea Entertainment Industry Association
• /
• v.14 no.4
• /
• pp.67-75
• /
• 2020

Metonymy can be said to be the only language's meaning shifting technique that exists in the domain of a single human thought in order to obtain a transparent cognitive effect. The purpose of this study was to analyze the 'descriptive metonymy' of the advertising content language constructed by the cognitive principle and to find a way to use it in media language education for social and cultural interests and reflection of college students. The metonymy used in advertising media contrasts with the difficulty of the metaphorical interpretation of "opaque and distant" reasoning. Storyboards, mostly focused on human emotions and behaviors, used metonymy's 'transparent and easy meaning shifting technique'. I have found that I can expect the efficiency of media language education that contains the interest and sociocultural interest, self-reflection, and future imagination of college students. Now, there is less need to perform cognitive reasoning for advertisements with ambiguous metaphor techniques. Lastly, in order to produce successful advertising content, we expect to use the language technique of 'narrative metonymy' with warm feelings of humans, and acknowledge the lack of quantitative research and leave it as a task for the next research.

• Science of Emotion and Sensibility
• /
• v.18 no.1
• /
• pp.3-14
• /
• 2015

An ad endorser's gaze direction is a salient nonverbal cue that consumers use in responding to advertisements. The gaze direction influences consumers' social perceptions (e.g., attractiveness, credibility) of the endorser and advertising effectiveness (e.g., advertising attitudes, brand attitudes). Especially, the cerebral emotional asymmetry hypothesis suggests that an ad endorser's left-averted gaze can produce more positive social perceptions and advertising effectiveness than the right-averted gaze for right-handed consumers. This study examined the effects of three gaze directions (direct, left-averted and right-averted gaze directions) of unknown female ad endorser on Korean males' advertising responses (attractiveness-, credibility- and ad-effectiveness-related responses), using online experimental method. The results indicated that the ad endorser's direct gaze was more likely to increase both positive (correspondence bias) and negative (suspicion, deceptiveness) social perceptions of her than the right-averted gaze. The direct gaze also created more positive advertising effectiveness (advertising attitudes) than the right-averted gaze. However, the study failed to find consistent differences in responses between left-averted gaze and either direct or right-averted gaze.

• Science of Emotion and Sensibility
• /
• v.16 no.1
• /
• pp.95-106
• /
• 2013

The study compares accuracies between humans and computer algorithms in the discrimination of spontaneous smiles from posed smiles. For this purpose, subjects performed two tasks, one was judgment with single pictures and the other was judgment with pair comparison. At the task of judgment with single pictures, in which pictures of smiling facial expression were presented one by one, subjects were required to judge whether smiles in the pictures were spontaneous or posed. In the task for judgment with pair comparison, in which two kinds of smiles from one person were presented simultaneously, subjects were to select spontaneous smile. To calculate the discrimination algorithm accuracy, 8 kinds of facial features were used. To calculate the discriminant function, stepwise linear discriminant analysis (SLDA) was performed by using approximately 50 % of pictures, and the rest of pictures were classified by using the calculated discriminant function. In the task of single pictures, the accuracy rate of SLDA was higher than that of humans. In the analysis of accuracy on pair comparison, the accuracy rate of SLDA was also higher than that of humans. Among the 20 subjects, none of them showed the above accuracy rate of SLDA. The facial feature contributed to SLDA effectively was angle of inner eye corner, which was the degree of the openness of the eyes. According to Ekman's FACS system, this feature corresponds to AU 6. The reason why the humans had low accuracy while classifying two kinds of smiles, it appears that they didn't use the information coming from the eyes enough.

• Science of Emotion and Sensibility
• /
• v.16 no.4
• /
• pp.535-544
• /
• 2013

Prior research on attribution has paid much attention to cognitive processes through which people infer causes. However, surprising omission from this research stream is the role of unconscious priming of different causes in task satisfaction and re-performance intention. This research is conducted to examine how attributing to different causes (chance vs. effort) via priming procedure affects task satisfaction and re-performance intention. To do this, participants were assigned to either chance-priming or effort-priming attribution condition and asked to answer O/X quiz (Exp 1) or multiple choice questions (Exp 2). Then, we provided 'Good' or 'Bad' feedback to participants randomly regardless of their actual task results. Finally, we measured their task satisfaction and re-performance intention. Results indicate that task satisfaction is affected by the valence of feedback they received but re-performance intention is influenced by interaction between attribution priming and the valence of feedback. Specifically, when receiving 'Good' feedback, participants in effort-priming condition have higher re-performance intention than those in chance-priming condition, whereas when receiving 'Bad' feedback, participants in effort-priming condition have lower re-performance intention than those in chance-priming condition. Implications of the findings are discussed in relation to the research for induction and decision making as well as psychology of addiction.

• Science of Emotion and Sensibility
• /
• v.25 no.2
• /
• pp.11-22
• /
• 2022

As social beings, we need to understand others' actions as quickly and accurately as possible. Action understanding can occur at many levels. We sometimes grasp others' intentions unintentionally. Other times, however, we have to expend effort to draw inferences about their goals. In the context of joint action, the joint Simon effect demonstrates that actors are influenced by the unintended representation of a co-actor's actions. This effect has been described as quasi-automatic, but it is unclear if the effect is automatic enough to be immune to cognitive load. Thus, we asked participants to complete a joint Simon task with or without a concurrent working memory task. One group of participants maintained a single digit in their mind during working memory load blocks (low-load group), while the other group maintained five digits (high-load group). As a result, the low-load group showed a joint Simon effect both during no-load and low-load blocks. In contrast, the high-load group had no joint Simon effect during either no-load or high-load blocks. These results suggest that the joint Simon effect is not an automatic phenomenon given that it requires cognitive resources. Actors in a joint task may represent a co-actor's actions in their task set, but only when cognitive resources are available.

• Journal of KIISE
• /
• v.44 no.7
• /
• pp.680-685
• /
• 2017

The emergence of online media and their data has enabled data-driven methods to solve challenging and complex tasks such as rumor classification problems. Recently, deep learning based models have been shown as one of the fastest and the most accurate algorithms to solve such problems. These new models, however, either rely on complete data or several days-worth of data, limiting their applicability in real time. In this study, we go beyond this limit and test the possibility of super early rumor detection via recurrent neural networks (RNNs). Our model takes in social media streams as time series input, along with basic meta-information about the rumongers including the follower count and the psycholinguistic traits of rumor content itself. Based on analyzing millions of social media posts on 498 real rumors and 494 non-rumor events, our RNN-based model detected rumors with only 30 initial posts (i.e., within a few hours of rumor circulation) with remarkable F1 score of 0.74. This finding widens the scope of new possibilities for building a fast and efficient rumor detection system.

• Journal of Digital Convergence
• /
• v.19 no.2
• /
• pp.397-402
• /
• 2021

The purpose of this study is to measure the satisfaction level of user experience of athlesure brand community by generation through surveys, and to derive satisfaction level according to consumption taste by generation. As a tool for measuring brand community, a questionnaire written Creating Pleasurable Interfaces Model of Stephen P. Anderson was used and was conducted focusing on women in their 10s-40s, which are the main consumers. As a result of the survey, the community participation rate and satisfaction level are significantly different depending on the consumption tendency of each generation. In addition, focused on task satisfaction was high, but focused on experience satisfaction was relatively low. It was found that there is a need for a plan to improve satisfaction based on experience for revitalization of online and offline communities in the future. Through this study, we expect the use of consumer-centered brand communities and brand communities according to the satisfaction level of consumption demand by generation and consumption preference by demand.