• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.2
• /
• pp.53-60
• /
• 2020

Sentiment analysis is the automated process of understanding attitudes and opinions about a given topic from written or spoken text. One of the sentiment analysis approaches is a dictionary-based approach, in which a sentiment dictionary plays an much important role. In this paper, we propose a method to automatically generate Korean sentiment lexicon from the well-known English sentiment lexicon called VADER (Valence Aware Dictionary and sEntiment Reasoner). The proposed method consists of three steps. The first step is to build a Korean-English bilingual lexicon using a Korean-English parallel corpus. The bilingual lexicon is a set of pairs between VADER sentiment words and Korean morphemes as candidates of Korean sentiment words. The second step is to construct a bilingual words graph using the bilingual lexicon. The third step is to run the label propagation algorithm throughout the bilingual graph. Finally a new Korean sentiment lexicon is generated by repeatedly applying the propagation algorithm until the values of all vertices converge. Empirically, the dictionary-based sentiment classifier using the Korean sentiment lexicon outperforms machine learning-based approaches on the KMU sentiment corpus and the Naver sentiment corpus. In the future, we will apply the proposed approach to generate multilingual sentiment lexica.

• Journal of Music and Human Behavior
• /
• v.13 no.2
• /
• pp.67-83
• /
• 2016

This study investigated the tonal characteristics in Korean emotion words by analyzing the pitch patterns transformed from word utterance. Participants were 30 women, ages 19-23. Each participant was instructed to talk about their emotional experiences using 4-syllable target words. A total of 180 utterances were analyzed in terms of the frequency of each syllable using the Praat. The data were transformed into meantones based on the semi-tone scale. When emotion words were used in the middle of a sentence, the pitch pattern was transformed to A3-A3-G3-G3 for '즐거워서(joyful)', C4-D4-B3-A3 for '행복해서(happy)', G3-A3-G3-G3 for '억울해서(resentful)', A3-A3-G3-A3 for '불안해서(anxious)', and C4-C4-A3-G3 for '침울해서(frustrated)'. When the emotion words were used at the end of a sentence, the pitch pattern was transformed to G4-G4-F4-F4 for '즐거워요(joyful)', D4-D4-A3-G3 for '행복해요(happy)', G3-G3-G3-A3 and F3-G3-E3-D3 for '억울해요(resentful)', A3-G3-F3-F3 for '불안해요(anxious)', and A3-A3-F3-F3 for '침울해요(frustrated)'. These results indicate the differences in pitch patterns depending on the conveyed emotions and the position of words in a sentence. This study presents the baseline data on the tonal characteristics of emotion words, thereby suggesting how pitch patterns could be utilized when creating a melody during songwriting for emotional expression.

• Journal of Korea Game Society
• /
• v.12 no.6
• /
• pp.131-144
• /
• 2012

To affect users' attitudes on social issues, some serious games use narrative which can be an effective tool to provoke users' emotions. In this paper, focus group interview on the experiences about playing was conducted to figure out if the parts of narratives containing educational goals can actually provoke certain emotions or not. From the interview, words having emotional meaning were extracted and matched with Korean Emotion Terms Database. Then the emotion terms were compared with Russell's schematic map of core affect and were categorized. The result showed that users mostly felt unpleasant emotions during the play. The unpleasant emotions helped to achieve the game's goal which is conveying the tragedy of war.

• Korean Journal of Cognitive Science
• /
• v.19 no.4
• /
• pp.477-497
• /
• 2008

Mood classification of blog text is an interesting problem, with a potential for a variety of services involving the Web. This paper introduces an approach to mood classification enhancements through the normalized negation n-grams which contain mood clues and corpus-specific term weighting(CSTW). We've done experiments on blog texts with two different classification methods: Enhanced Mood Flow Analysis(EMFA) and Support Vector Machine based Mood Classification(SVMMC). It proves that the normalized negation n-gram method is quite effective in dealing with negations and gave gradual improvements in mood classification with EMF A. From the selection of CSTW, we noticed that the appropriate weighting scheme is important for supporting adequate levels of mood classification performance because it outperforms the result of TF*IDF and TF.

• Science of Emotion and Sensibility
• /
• v.20 no.4
• /
• pp.31-40
• /
• 2017

This research aims to investigate 'Hallyu' contents consumption tendency of consumers from Korea, Japan, and the United States by analyzing their emotional responses. With the development of social media, research on emotion analysis by reviewing text materials has grown. Whereas environmental variables affect consumer demand towards 'Hallyu' contents, little comparative analyses have been conducted on the emotional responses of consumers from different countries. In this research, the emotional prototype model proposed by Russell(1980) used to extract and distinguish emotional words to clarify how people in the three countries differently perceive the Korean drama "Goblin". First of all, the SNS reviews were collected during a two-month period (February 12 to April 12). Second, significant factors were identified in the collected data according to Russell's emotion model. Third, random forest was applied to organize the selected variables in the order of variable importance. Fourth, the correlations among the emotional words were compared. Lastly, the accuracy of the trained model was measured using the test dataset. The results show that "Happy" was found to be the greatest factor in Korea and in the United States and "Pleased" in Japan. Emotional words correlations showed that when watching the drama "Goblin", "passive unpleasure" was the main factor associated with individual's interest in Korea whereas "passive pleasure" was associated with individual's interest in Japan and in the United States. Based on the results, this research suggests the possibility of developing evaluation guidelines for emotional responses of different countries towards 'Hallyu' contents.

• Journal of Korea Game Society
• /
• v.12 no.6
• /
• pp.95-106
• /
• 2012

Emotion is very important feature in educational situation. Because it has high influence to memory, educational achievement, motivation. This study tried to find out possibility of serious game as emotional engagement tool in educational situation. We did our pilot experiment to elementary school students who are english as second language. In this L2 learning situation, we did our basic experiment with English language learning serious game called 'Word Collectrian". Word Collectrian has some features for emotional engagement. It has interaction for dynamic word visualization, providing context video for word usage, putting visualized word on learner's virtual home. According to experimental result, word Collectrian has possibility for educational achievement and emotional engagement effect.

• Proceedings of the Korean Information Science Society Conference
• /
• 2012.06c
• /
• pp.51-53
• /
• 2012

데이터 마이닝의 문서분류 기술에서 발전된 오피니언 마이닝은 이제 국외뿐만 아니라 국내의 학계 및 기업에서 중요한 관심분야로 자리잡아가고 있다. 오피니언 마이닝의 핵심은 문서에서 감정 단어를 추출하여 긍정/부정 여부를 얼마나 정확하게 자동적으로 판별하느냐를 평가하는 것이다. 국내에서도 이에 관련된 많은 연구가 이루어 졌으나 아직 실용적으로 적용할 만큼의 정확한 분류 정확도 보이지 않고 있다. 그 이유는 한국어의 경우 비문법적 표현, 감정단어의 다양성 등으로 인해 문서의 극성을 판별하기가 쉽지 않기 때문이다. 본 논문에서는 문법적 요소를 최대한 배제하고 단어 패턴의 빈도만을 고려한 영화평 분류기법을 제안한다. 제안된 방법에서는 문서를 단어들의 리스트로 추상화하여 패턴들의 빈도로 학습한 후 적절한 스코어 함수를 적용하여 문서의 극성을 판별한다. 또한 실험을 통해 제안된 기법의 정확도를 평가한다.

• Annual Conference on Human and Language Technology
• /
• 2019.10a
• /
• pp.519-521
• /
• 2019

최근 소셜 메신저를 통해 많은 사람들이 의사소통을 주고받음에 따라, 텍스트에서 감정을 파악하는 것이 중요하다. 따라서, 감정이 태깅된 데이터가 필요하다. 하지만, 기존 연구는 감정이 태깅된 데이터의 양이 많지가 않다. 이는 텍스트에서 감정을 파악하는데 성능 저하를 야기할 수 있다. 이를 해결하기 위해, 본 논문에서는 단어 매칭 방법과 형태소 매칭 방법을 이용하여 많은 양의 한국어 감정 태깅 데이터셋인 EmoNSMC 를 구축하였다. 구축한 데이터셋은 네이버 영화 감상 리뷰 데이터 (NSMC)에 디스턴트 수퍼비전 방법 (distant supervision) 방법을 적용하여 weak labeling을 진행하였고, 이 과정에서 한국어 감정 어휘 사전 (KTEA) 을 이용하였다. 구축된 데이터셋의 감정 분포 결과, 형태소 매칭 방법을 통해 구축한 데이터셋이 좀 더 감정 분포가 균등한 것을 확인할 수 있었다. 해당 데이터셋은 공개되어 있다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2006.05a
• /
• pp.533-536
• /
• 2006

본 논문에서는 음성 신호로부터 감정의 특징을 나타내는 요소를 찾아내는 것을 목표로 하고 있다. 일반적으로 감정을 인식할 수 있는 요소는 단어, 톤, 음성신호의 피치, 포만트, 그리고 발음 속도 및 음질 등이 있다. 음성을 기반으로 감정을 익히는 방법 중에서 현재 가장 많이 접근하고 있는 방법은 피치에 의한 방법이 있다. 사람의 경우는 주파수 같은 분석 요소보다는 톤과 단어, 빠르기, 음질로 감정을 받아들이게 되는 것이 자연스러운 방법이므로 이러한 요소들이 감정을 분류하는데 중요한 요소로 쓰일 수 있다. 따라서, 본 논문에서는 감정에 따른 음성의 특징을 추출하기 위해 사람의 감정 중에서 비교적 자주 쓰이는 평상, 기쁨, 화남, 슬픔에 관련된 4가지 감정을 비교 분석하였으며, 인간의 감정에 대한 음성의 특성을 분석한 결과, 강도와 스펙트럼에서 각각의 일관된 결과를 추출할 수 있었고, 이러한 결과에 대한 실험 과정과 최종 결과 및 근거를 제시하였다. 끝으로 실험에 의해 제안한 방법의 유용성을 입증하고자 한다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2013.07a
• /
• pp.31-32
• /
• 2013

본 논문에서는 영화의 등장인물의 성향을 파악하기 위해 시나리오의 대사로부터 감정어를 추출하고, 등장인물의 감정어들을 긍정, 부정, 중립의 3개로 단순화하여 등장인물의 성향을 가시화 시켜주는 방법을 제안한다. 대사로부터 감정어를 추출하기 위해 WordNet 기반의 감정어 추출 방법을 제안한다. WordNet은 단어 간에 상위어와 하위어, 유사어 등의 관계로 연결된 네트워크 구조의 사전이다. 이 네트워크 구조에서 최상위의 감정 항목과의 거리를 계산하여 단어별 감정량을 계산하여 대사를 30 차원의 감정 벡터로 표현한다. 등장인물별로 추출된 감정 벡터를 긍정, 부정, 중립의 3개의 차원으로 단순화 하여 등장인물의 성향을 표현한다.