• Proceedings of the Korean Society of Computer Information Conference
• /
• 2010.07a
• /
• pp.411-412
• /
• 2010

전통적으로 자연어 텍스트로부터의 감정인식 연구는 감정 키워드에 기반한다. 그러나 감정 키워드만을 이용하면 자연어 문장이 원래 갖고 있는 통사정보나 의미정보는 잃어버리게 된다. 이를 극복하기 위해 본 논문에서는 자연어 텍스트를 Predicate Logic 형태로 변환하여 감정 정보처리의 기반데이터로 사용한다. Predicate Logic형태로 변환하기 위해서 의존 문법 구문분석기를 사용하였다. 이렇게 생성된 Predicate 데이터 중 감정 정보를 갖고 있는 Predicate만을 찾아내는데 이를 위해 Emotional Predicate Dictionary를 구축하였고 이 사전에는 하나의 Predicate마다 미리 정의된 개념 클래스로 사상 시킬 수 있는 정보를 갖고 있다. 개념 클래스는 감정정보를 갖고 있는지, 어떤 감정인지, 어떤 상황에서 발생하는 감정인지에 대한 정보를 나타낸다. 자연어 텍스트가 Predicate으로 변환되고 다시 개념 클래스로 사상되고 나면 KBANN으로 구현된 Lazarus의 감정 생성 규칙에 적용시켜 최종적으로 인식된 감정을 판단한다. 실험을 통해 구현된 시스템이 인간이 인식한 감정과 약 70%이상 유사한 인식 결과를 나타냄을 보인다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2017.10a
• /
• pp.506-509
• /
• 2017

Opinion mining is a way to analyze the emotions in the text and is used to identify the emotional state of the author and to find out the opinions of the public. As you can analyze individual emotions through opinion mining, if you analyze the text by region, you can find out the emotional state you have in each region. The regional sentiment analysis can obtain information that could not be obtained from personal sentiment analysis, and if a certain area has emotions, it can understand the cause. For regional sentiment analysis, we need text data created by region, so we need to collect data through Twitter crawling. Therefore, this paper designs a Twitter data collection system for regional sentiment analysis. The client requests the tweet data of the specific region and time, and the server collects and transmits the requested tweet data from the client. Through the latitude and longitude values of the region, it collects the tweet data of the area, and it can manage the text by region and time through collected data. We expect efficient data collection and management for emotional analysis through the design of this system.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.9 no.6
• /
• pp.681-689
• /
• 2019

The 46th Davos Forum of the World Economic Forum (WEF) predicts the continued growth of the 4th industry in the future. Currently, the 4th industry is attracting attention in various academic and practical fields. As a core technology of the 4th industry, Big Data is regarded as a major resource to lead the 4th industrial revolution along with artificial intelligence. As the growing interest in Big Data, researches on it are actively being done. However, literature studies on existing Big Data are focused on qualitative research, and quantitative research is insufficient. Therefore, this study aims to analyze the big data research flow in MIS field and to make academic thirst for quantification. This study has collected 145 abstracts of big data papers published in major journals in MIS field and confirmed that a majority of papers are published in Decision Support Systems Journal. Text mining and text network analysis were performed only for DSS journals to eliminate bias. As a result of the analysis, it was found out that researches on combining big data in the management field between 2012 and 2014, and researches on system development and analysis method for using big data from 2015 to 2017 were conducted.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.11
• /
• pp.481-492
• /
• 2023

This study explores how to build a Korean dataset to extract information from text using generative large language models. In modern society, mixed information circulates rapidly, and effectively categorizing and extracting it is crucial to the decision-making process. However, there is still a lack of Korean datasets for training. To overcome this, this study attempts to extract information using text-based zero-shot learning using a generative large language model to build a purposeful Korean dataset. In this study, the language model is instructed to output the desired result through prompt engineering in the form of "system"-"instruction"-"source input"-"output format", and the dataset is built by utilizing the in-context learning characteristics of the language model through input sentences. We validate our approach by comparing the generated dataset with the existing benchmark dataset, and achieve 25.47% higher performance compared to the KLUE-RoBERTa-large model for the relation information extraction task. The results of this study are expected to contribute to AI research by showing the feasibility of extracting knowledge elements from Korean text. Furthermore, this methodology can be utilized for various fields and purposes, and has potential for building various Korean datasets.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2017.04a
• /
• pp.837-840
• /
• 2017

각종 사이버 범죄가 증가함에 따라 실시간 모니터링을 통한 사전 탐지 기술뿐만 아니라, 사후 원인 분석을 통한 사고 재발 방지 기술의 중요성이 증가하고 있다. 사후 분석은 시스템에서 생산된 다양한 유형의 대용량 로그를 기반으로 분석가가 보안 위협 과정을 규명하는 것으로 이를 지원하는 다양한 상용 및 오픈 소스 SW 존재하나, 대부분 단일 분석가 PC에서 운용되는 파일 기반 SW로 대용량 데이터에 대한 분석 성능 저하, 다수 분석가 간의 데이터 공유 불가, 통계 연관 분석 한계 및 대화형 점진적 내용 분석 불가 등의 문제점을 해결하지 못하고 있다. 이러한 문제점을 해결하기 위하여 고성능 인메모리 관계형 데이터베이스 시스템을 로그 스토리지로 활용하는 대용량 로그 분석 SW 개발하였다. 특히, 기 확보된 공격자 프로파일을 활용하여 공격의 유무를 확인하는 텍스트 패턴 매칭 연산은 전통적인 관계형 데이터베이스 시스템의 FTS(Full-Text Search) 기능 활용이 가능하나, 대용량 전용 색인 생성에 따른 비현실적인 DB 구축 소요 시간과 최소 3배 이상의 DB 용량 증가로 인한 시스템 리소스 추가 요구 등의 단점이 있다. 본 논문에서는 인메모리 관계형 데이터베이스 시스템 기반 효율적인 텍스트 패턴 매칭 연산을 위하여, 고성능의 대용량 로그 DB 적재 방법과 새로운 유형의 패턴 매칭 방법을 제안하였다.

• The Journal of the Korea Contents Association
• /
• v.8 no.11
• /
• pp.10-17
• /
• 2008

The sign recognition and translation offer information and support decision making for foreigners or city tourist. Collecting sign images and building words in signs are essential to train machine recognizers and to evaluate systems. In this paper, we analyze the characteristics of sign images. The collected sign images are about 1000 captured from difference conditions and locations. We also build a dictionary of words in 100,000 sign names.

• The Journal of the Korea Contents Association
• /
• v.18 no.6
• /
• pp.217-234
• /
• 2018

In this study, big data analysis was conducted for domestic and international sports goods brands. Text Mining, TF-IDF, Opinion Mining, interestity graph were conducted through the social matrix program Textom and the fashion data analysis platform MISP. In order to examine the recent recognition of sports brands, the period of study is limited to 1 year from January 1, 2017 to December 31, 2017. As a result of analysis, first, we could confirm the products representing each brand. Second, I could confirm the marketing that represents each brand. Third, the common words extracted from each brand were identified. Fourth, the emotions of positive and negative of each brand were confirmed.

• Journal of the Korea Society of Computer and Information
• /
• v.25 no.2
• /
• pp.123-131
• /
• 2020

The text ranking algorithm is a representative method for keyword extraction, and its importance is emphasized highly. In this paper, we compare the performance of recent research and experiments with TF-IDF, SMART, INQUERY and CCA algorithms, which are used in text ranking algorithm.. After explaining each algorithm, we compare the performance of each algorithm based on the data collected from news and Twitter. Experimental results show that all of four algorithms can extract specific words from news data equally. However, in the case of Twitter, CCA has the best performance to extract specific words, and INQUERY shows the worst performance. We also analyze the accuracy of the algorithm through six comparison metrics. The experimental results present that CCA shows the best accuracy in the news data. In case of Twitter, TF-IDF and CCA show similar performance and demonstrate good performance.

• Journal of Korea Game Society
• /
• v.22 no.1
• /
• pp.87-98
• /
• 2022

With the development of fourth industrial revolution-related technology and increased demands for non-face-to-face services, VR games attract attention. This study collected VR game review data from an online game platform STEAM and analyzed chronical trends using text mining and word cloud analysis. According to the results, experience and perceived cost were major trends from 2016 to 2017, increased demands for FPS and rhythm games were from 2018 to 2019, and story and immersion were from 2020 to 2021. It aims to contribute to expanding the base of VR games by identifying the keywords VR users take interest in by period.

• Annual Conference on Human and Language Technology
• /
• 2022.10a
• /
• pp.251-256
• /
• 2022

범주 불균형은 분류 모델이 다수 범주에 편향되게 학습되어 소수 범주에 대한 분류 성능을 떨어뜨리는 문제를 야기한다. 언더 샘플링 기법은 다수 범주 데이터의 수를 줄여 소수 범주와 균형을 이루게하는 대표적인 불균형 해결 방법으로, 텍스트 도메인에서의 기존 언더 샘플링 연구에서는 단어 임베딩과 랜덤 샘플링과 같은 비교적 간단한 기법만이 적용되었다. 본 논문에서는 트랜스포머 기반 문장 임베딩과 군집화 기반 샘플링 방법을 통해 텍스트 데이터의 정보 손실을 최소화하는 언더샘플링 방법을 제안한다. 제안 방법의 검증을 위해, 감성 분석 실험에서 제안 방법과 랜덤 샘플링으로 추출한 훈련 세트로 모델을 학습하고 성능을 비교 평가하였다. 제안 방법을 활용한 모델이 랜덤 샘플링을 활용한 모델에 비해 적게는 0.2%, 많게는 2.0% 높은 분류 정확도를 보였고, 이를 통해 제안하는 군집화 기반 언더 샘플링 기법의 효과를 확인하였다.