• Korean Journal of Cognitive Science
• /
• v.28 no.4
• /
• pp.193-221
• /
• 2017

Grouping episodes into semantically related categories is necessary for better mnemonic structure. However, the effect of grouping on memory of subordinate details was not clearly understood. In an fMRI study, we tested whether attending superordinate during semantic association disrupts or enhances subordinate episodic details. In each cycle of the experiment, five cue words were presented sequentially with two related detail words placed underneath for each cue. Participants were asked whether they could imagine a category that includes the previously shown cue words in each cycle, and their confidence on retrieval was rated. Participants were asked to perform cued recall tests on presented detail words after the session. Behavioral data showed that reaction times for categorization tasks decreased and confidence levels increased in the third trial of each cycle, thus this trial was considered to be an important insight where a semantic category was believed to be successfully established. Critically, the accuracy of recalling detail words presented immediately prior to third trials was lower than those of followed trials, indicating that subordinate details were disrupted during categorization. General linear model analysis of the trial immediately prior to the completion of categorization, specifically the second trial, revealed significant activation in the temporal gyrus and inferior frontal gyrus, areas of semantic memory networks. Representative Similarity Analysis revealed that the activation patterns of the third trials were more consistent than those of the second trials in the temporal gyrus, inferior frontal gyrus, and hippocampus. Our research demonstrates that semantic grouping can cause memories of subordinate details to fade, suggesting that semantic retrieval during categorization affects the quality of related episodic memory.

• Journal of the Korean Society for information Management
• /
• v.10 no.1
• /
• pp.97-124
• /
• 1993

This study was undertaken in order to develop practical automatic indexing techniques suitable for Korean natural language texts. It has taken a modest step toward this goal by developing an automatic syntactico-statistical indexing method and evaluating the method by comparing the resutls with manual indexing. For this experimental study, the Korean text database was constructed manually based on 300 abstracts covering business subject. The experimental results showed that the performance of the automatic syntactico-statistical indexing system was comparable to that of other studies which have compared automatic indexing with manual indexing.

• 한국어정보학회:학술대회논문집
• /
• 2017.10a
• /
• pp.333-337
• /
• 2017

본 논문에서는 개체명 인식과 언어 모델의 다중 학습을 이용한 한국어 개체명 인식 방법을 제안한다. 다중 학습은 1 개의 모델에서 2 개 이상의 작업을 동시에 분석하여 성능 향상을 기대할 수 있는 방법이지만, 이를 적용하기 위해서 말뭉치에 각 작업에 해당하는 태그가 부착되어야 하는 문제가 있다. 본 논문에서는 추가적인 태그 부착 없이 정보를 획득할 수 있는 언어 모델을 개체명 인식 작업과 결합하여 성능 향상을 이루고자 한다. 또한 단순한 형태소 입력의 한계를 극복하기 위해 입력 표상을 자소 및 형태소 품사의 임베딩으로 확장하였다. 기계 학습 방법은 순차적 레이블링에서 높은 성능을 제공하는 Bi-directional LSTM CRF 모델을 사용하였고, 실험 결과 언어 모델이 개체명 인식의 오류를 효과적으로 개선함을 확인하였다.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.25-30
• /
• 2023

초거대 언어모델은 과연 수수께끼 문제에 재치있는 답변을 할 수 있을까? 최근 초거대 언어모델(Large language model, LLM)은 강력한 성능 및 유저 만족도를 보이며 세계의 이목을 집중시키고 있다. 여러 태스크들에 대한 정량 평가를 넘어서 최근에는 LLM의 창의력 및 고도화된 언어능력을 평가하는 연구들이 등장하고 있다. 본 논문에서는 이러한 연구 흐름에 따라 LLM의 재치에 관해 고찰해본다. 이때 재치를 평가하기 위한 태스크로 이를 요구하는 말놀이인 수수께끼를 활용한다. 본 논문은 LLM이 수수께끼를 잘 수행하는지를 모델 추론을 통해 평가하며, 모델 추론 시 활용되는 프롬프트들의 성격에 따른 성능 변화를 관찰한다. 또한 수수께끼의 종류에 따른 모델의 능력을 비교 분석하며 LLM의 추론 결과에 대한 오류 분석을 수행한다. 본 논문은 실험을 통해 GPT-4가 가장 높은 성능을 보이며, 설명글이나 데이터 예시를 추가할 시 성능을 한층 더 향상시킬 수 있음을 확인한다. 또한 단어 기반보다는 특성 기반의 수수께끼에 더욱 강력한 성능을 보이며, 오류 유형 분석을 통해 LLM이 환각(hallucination) 문제와 창의력을 동시에 가지고 있다고 분석한다.

• The Journal of Korean Institute of Next Generation Computing
• /
• v.13 no.5
• /
• pp.7-18
• /
• 2017

Recently, the number of SNS user has rapidly increased due to smart device industry development and also the amount of generated data is exponentially increasing. In the twitter, Text data generated by user is a key issue to research because it involves events, accidents, reputations of products, and brand images. Twitter has become a channel for users to receive and exchange information. An important characteristic of Twitter is its realtime. Earthquakes, floods and suicides event among the various events should be analyzed rapidly for immediately applying to events. It is necessary to collect tweets related to the event in order to analyze the events. But it is difficult to find all tweets related to the event using normal keywords. In order to solve such a mentioned above, this paper proposes A Generation and Matching Method of Normal-Transient Dictionary for realtime topic detection. Normal dictionaries consist of general keywords(event: suicide-death-loop, death, die, hang oneself, etc) related to events. Whereas transient dictionaries consist of transient keywords(event: suicide-names and information of celebrities, information of social issues) related to events. Experimental results show that matching method using two dictionary finds more tweets related to the event than a simple keyword search.

• Korean Journal of Child Education & Care
• /
• v.17 no.4
• /
• pp.233-258
• /
• 2017

Although listening to music is the most basic musical experience that is ahead of all the other musical activities, it tends to be neglected due to lack of awareness and difficult teaching methods. This study is to compose integrative music appreciation by reflecting thoughts of children, draw children's attention to listening to music, and let them to discover and create musical concept and structures by themselves, accompanied by related activities through which children can express their thoughts and feelings with children's poem. Considering the peculiarities of three-four year-olds, it may be difficult for them to create poems individually, so we let them to write poems through small group discussion in which they can share their opinions and observe their peers' reaction. We provided a teaching method to teachers who find activities of listening to music and writing poems difficult, then we analyzed the effect. When children finished writing poems in small groups through integrative music appreciation reflecting their thoughts, they placed greater weight on preparing to listen to music at first, but going through the program, they discovered musical concepts and became active in music appreciation. In the related activity, writing poems inspired them to think creatively, listening to their peer's stories. Even children who were not interested in children's poem showed higher participation. Teachers found children's creative words to be interesting, discovering the joy of creation.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.05a
• /
• pp.447-450
• /
• 2019

본 논문은 한국어 문장의 감성 분류를 위해 문장의 형태소, 음절, 자소를 입력으로 하는 합성곱층과 DenseNet 을 적용한 Text Multi-channel DenseNet 모델을 제안한다. 맞춤법 오류, 음소나 음절의 축약과 탈락, 은어나 비속어의 남용, 의태어 사용 등 문법적 규칙에 어긋나는 다양한 표현으로 인해 단어 기반 CNN 으로 추출 할 수 없는 특징들을 음절이나 자소에서 추출 할 수 있다. 한국어 감성분석에 형태소 기반 CNN 이 많이 쓰이고 있으나, 본 논문에서 제안한 Text Multi-channel DenseNet 모델은 형태소, 음절, 자소를 동시에 고려하고, DenseNet 에 정보를 밀집 전달하여 문장의 감성 분류의 정확도를 개선하였다. 네이버 영화 리뷰 데이터를 대상으로 실험한 결과 제안 모델은 85.96%의 정확도를 보여 Multi-channel CNN 에 비해 1.45% 더 정확하게 문장의 감성을 분류하였다.

• Journal of Korean language education
• /
• v.29 no.2
• /
• pp.25-48
• /
• 2018

The aim of this study was to analyze research trends in the fields of Korean Language and Literature, Korean Language Education, and Korean Language Education as a Foreign Language. To this end, key words were extracted from 29,470 academic theses published between 2000 and 2017. The results of the analysis are as follows. First, in the field of Korean Language and Literature, there is greater quantity in studies about Korean language than about literature, and starting from 2010, there was an increase in studies similar to those from the field of Korean Language Education as a Foreign Language. Next, in comparison to the other fields, the field of Korean Language Education has greater variance in its research theme-in particular, numerous studies related to the site of education. Finally, the field of Korean Language Education has the following trends: a) there are copious studies focused on Korean language learners in comparison to other fields, b) there are a greater number of studies focused on culture, and c) the key words change by time period which suggest that research demands transformed over time. Overall, a total of 64 highest frequency key words from the three academic fields were investigated. Of these, 22 were common key words and 42 were differential key words. In this way, it was possible to illuminate the identity of each field.

• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.1
• /
• pp.13-18
• /
• 2019

In this study, we investigate important keywords and their relationships among the keywords for social issues, and analyze topics to find subjects of the social issues. In particular, we collected twitter data with the keyword 'metoo' which has attracted much attention in these days, and perform keyword analysis and topic modeling. First, we preprocess the twitter data, identified important keywords, and analyzed the relatedness of the keywords. After then, topic modeling is performed to find subjects related to 'metoo'. Our experimental results showed that relatedness of keywords and subjects on social issues in twitter are well identified based on keyword analysis and topic modeling.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.05a
• /
• pp.496-498
• /
• 2021

In Korea, based on the high fertility rate before 1980, the total population has been steadily increasing, and since the mid-1980s, the fertility rate has fallen sharply and has fallen below the level of population replacement. The cause of low birth rate in the region is not voluntary rejection, but rather, it is necessary to find out the cause by identifying the structural causes of the local community from various angles. We collected local Internet news and local representative cafe data, where many mothers participate, based on the budget area with a very low fertility rate among various areas. Factors of childbirth inhibition were analyzed by using the frequency of concurrent words that became issues related to population decline, low birthrate, and child-rearing welfare.