Earticle

인터넷 기술과 다양한 정보 기술의 융합을 통해 광대한 정보를 수집 및 가공하여 사용자 취향에 맞는 다양한 지식을 서로 주고받을 수 있게 되었다. 특히, 영화 미디어 속에 내포된 감정 변화의 흐름을 통해 사용자의 취향과 연결된 밀접한 콘텐츠를 선호하는 경향이 있다. 사용자는 스크립트에 나타난 정보를 바탕으로 전체적인 감정의 흐름이나 특정한 등장인물, 또는 특정한 장면에서의 감정의 흐름을 보다 빠르게 이해하기 위해 시각화하고자 한다. 본 논문에서는 영화 웹페이지로부터 정제되지 않은 데이터를 획득한 후, 정제 과정을 거쳐 표준화된 스크립트 형식으로 변환한다. 정제된 데이터를 XML 문서로 변환하여 다양한 정보를 쉽게 획득한 후, 감정 예측 시스템에 각 문단을 입력하여 다양한 감정을 예측한다. 예측된 감정들의 흐름과 스크립트에 포함된 정보량을 혼합하여 사용자가 필요로 하는 다양한 감정의 전체적 또는 특정 부분에서 캐릭터들 간의 감정 상태의 변화를 시각화를 통해 쉽게 이해할 수 있는 시스템을 제안한다.

Through the convergence of Internet technology and various information technologies, it is possible to collect and process vast amount of information and to exchange various knowledge according to user's personal preference. Especially, there is a tendency to prefer intimate contents connected with the user's preference through the flow of emotional changes contained in the movie media. Based on the information presented in the script, the user seeks to visualize the flow of the entire emotion, the flow of emotions in a specific scene, or a specific scene in order to understand it more quickly. In this paper, after obtaining the raw data from the movie web page, it transforms it into a standardized scenario format after refining process. After converting the refined data into an XML document to easily obtain various information, various sentences are predicted by inputting each paragraph into the emotion prediction system. We propose a system that can easily understand the change of the emotional state between the characters in the whole or a specific part of the various emotions required by the user by mixing the predicted emotions flow and the amount of information included in the script.

18th of May means the memorial (symbolic) day for Kwang-Ju citizens who resisted against Korean army in order to democratize South Korea in 1980. <Fig.1> illustrates the still image of 5.18 Kwang-Ju affairs. This work was shown in Hub City of Asian Culture Project, “Integrating”, administered by the national government of Korea in Kwang-Ju, 2007, for memorial of 5.18 Kwang-Juaffairs[1]. As the proverb says, “It's better to see it once than hearing it a thousand times.” Eyes as the visual organ play an important role in order to operate human sense organ. These visual images of 5.18 affairs are aroused from their resistance souls and reported for the live scene in those days. Indirectly we are able to feel an emotion of tremble like a fear of 5.18. Inspiration of visual scene propagates through human brain-waves.

The purpose of this study is to investigate emotional changes in an Act and across Acts and the emotion relationship between major characters in Shakespeare's four great tragedies through visualization of sentiment analysis using dialogue corpora, a VADER sentiment word-list, Python 3.5, and Netminer 4.4. We use both bar graphs and line charts as time visualization. The bar charts show how emotional changes between major characters are compared within an Act and across Acts in the normalized positive and negative sentiment valence sum. The line charts show who has emotional ups and downs throughout the whole play in the average sentiment valence. This study also attempts to visualize network structures connecting major characters and discuss sentiment intensity to tell us how much positive or negative sentiment there is in their detailed emotion relationship, representing relation visualization.

Despite advances in Natural Language Processing (NLP) and Large Language Models (LLMs), developing a CBT chatbot that provides emotional support and fosters continuous engagement remains a challenging task. To address this, we proposed SupportlyChat, which is a fine-tuned LLaMA with a custom dataset and provides emotion analysis visualization. We curated the custom dataset by integrating mental health conversation dataset and HappyDB dataset which collects people’s happy moments. We conducted a user study with 10 participants using conditions including standard chatbot responses, emotion analysis reports with standard responses, and emotion analysis reports with empathetic responses. The findings indicate that participants found empathetic responses helpful in shifting towards positive thinking with our proposed method. Additionally, they preferred emotional visualization for its ability to enhance professionalism and recognize inherent emotions. Based on these findings, we suggest utilizing empathetic responses in psychological counseling to emotionally support clients and using emotion analysis reports to promote continuous engagement.

본 연구는 전화 통화를 부담스러워하고 텍스트 기반 소통을 선호하는 이른바 '콜포비아' 경향에 주목하여, AI챗봇 기반 감정 시각화 서비스가 사용자의 심리적 편안함과 사용자 애착 형성에 미치는 영향을 탐색하였다. 설문조사(n=61)와 심층인터뷰(n=5)를 병행한 혼합 연구를 수행하였으며, Peter Morville의 UX 허니콤 모델을 기반으로 한 설문과 Jakob Nielsen의 반구조화 인터뷰를 활용하였다. 수집된 데이터는 비모수 통계 분석과 주제 코딩을 통해 분석하였다. 연구 결과, AI의 비판단성이 감정 표현의 심리적 장벽을 낮추는 핵심 요인으로 나타났으며(Q9a, M=3.98), 시각화된 감정 기록과 누적 구조는 사용자에게 성취감과 사용자 애착을 형성하였다. 허니콤 7요소 분석에서는 가치성(Valuable, M=4.14)이 가장 높게 나타났으며, 캘린더 기반 감정 시각화의 의미(Q16, M=4.23)가 개별 문항 최상위를 기록하였다. AI 사용 경험이 있는 집단은 그렇지 않은 집단에 비해 사용성과 탐색 가능성에서 유의미하게 높은 점수를 보였다(p<.05). 이는 콜포비아 성향과 AI 감정 발화의 연결을 실증한 탐색적 연구로서 의의를 갖는다. 이러한 결과는 비판단적 상호작용과 시간 기반 누적 시각화가 AI 기반 감정 서비스 설계의 핵심 원칙으로 기능할 수 있음을 시사한다.

This study investigates "call phobia" ? a preference for text-based communication over phone calls ? and explores how an AI chatbot-based emotion visualization service influences users' psychological comfort and user attachment. A mixed-methods design was employed, combining a survey (n=61) based on Peter Morville’s Honeycomb UX model and semi-structured interviews (n=5) informed by Jakob Nielsen’s principles, with data analyzed through non-parametric statistics and thematic coding. The results indicate that AI's non-judgmental nature lowers barriers to emotional expression (Q9a, M=3.98), while visualized emotional data and their temporal accumulation foster user attachment. Among the Honeycomb dimensions, Valuable scored highest (M=4.14), and the perceived meaningfulness of calendar-based emotion visualization (Q16, M=4.23) was the highest-rated individual item. Users with prior AI experience showed significantly higher usability and findability scores (p<.05). > This study empirically links call phobia to AI-based emotional expression. These findings suggest that non-judgmental interaction and temporal accumulation serve as key design principles for AI-based emotion-support services.

A convolutional neural network (CNN) has been widely used in facial expression recognition (FER) because it can automatically learn discriminative appearance features from an expression image. To make full use of its discriminating capability, this paper suggests a simple but effective method for CNN based FER. Specifically, instead of an original expression image that contains facial appearance only, the expression image with facial geometry visualization is used as input to CNN. In this way, geometric and appearance features could be simultaneously learned, making CNN more discriminative for FER. A simple CNN extension is also presented in this paper, aiming to utilize geometric expression change derived from an expression image sequence. Experimental results on two public datasets (CK+ and MMI) show that CNN using facial geometry visualization clearly outperforms the conventional CNN using facial appearance only.

본 논문은 음악의 감성을 시각화하는 새로운 방법을 제안하는 논문으로 시각화를 위하여 감성 LED조명 제어 시스템을 고안하였으며 이 시스템은 청중들의 음악 경험을 더욱 풍요롭게 하기 위함에 목적이 있다. 이 시스템은 기존 연구의 음악 감성 분석 알고리즘을 이용하였는데 하나의 음악에서 하나의 감성을 추출하는 기존 연구의 한계를 보완하고자 음악의 1초당 세그먼트에서 감성을 추출 하여 감성 변화를 감지하는 방법을 연구하였다. 또한 LED조명의 감성 컬러는 IRI컬러 모델과 Thayer의 감성 모델을 기초로 새로운 감성 컬러 36가지를 정의 하여 각 감성에 맞는 색으로 연출 할 수 있도록 구성하고 애니메이션과 더불어 감성 LED 조명으로 표현 할 수 있도록 하였다. 이 시스템의 검증을 위한 실험 결과, 듣는이의 감성과 감성 LED 조명 시스템의 감성 일치율이 60% 이상으로 향후 음악 감성 기반 서비스의 중요한 밑거름이 될 것으로 보인다.

This paper proposes a new strategy of emotion-based music visualization. Emotional LED lighting control system is suggested to help audiences enhance the musical experience. In the system, emotion in music is recognized by a proposed algorithm using a dimensional approach. The algorithm used a method of music emotion variation detection to overcome some weaknesses of Thayer's model in detecting emotion in a one-second music segment. In addition, IRI color model is combined with Thayer's model to determine LED light colors corresponding to 36 different music emotions. They are represented on LED lighting control system through colors and animations. The accuracy of music emotion visualization achieved to over 60%.

This paper introduces one methodology of Japanese research based on data and aims to explore the co-occurrence relationship of Japanese emotion vocabulary using “network analysis”.　The subject of the analysis is “100 books of Shincho Bunko”. Specifically, it collates texts of 65 Japanese novels in “100 of Shincho Bunko”, which completed the morphological comprehensive analysis, to “Japanese sentiment expression dictionary (JIWC-Dictionary)” which was used as a criterion for categorizing emotions. Targeting two-character kanji noun vocabulary, it then uses the network analysis method and visualizes the co-occurrence relationship of the various vocabulary that belong to the emotion categories of “sadness”, “anxiety”, “anger”, “disgust”, “trust”, “surprise”, and “excited”.　 The two-character kanji noun vocabulary to be analyzed was 33,207 out of 134,370 total tokens, of which 211 were output as “TRUE” from the seven emotion categories.　“Anxiety” shows the highest number with 45, followed by “trust” 36, “disgust” 30, “sadness” 26, “surprise” and “excited” each 25, and 24 from “anger”.　With these results, we observed the aspect of the vocabulary by emotion category, and at the same time, listed the co-occurrence relationship for the “earthquake” of “anxiety”, which contained the most vocabulary in the emotion category.