• Journal of the Korean Society for information Management
• /
• v.32 no.1
• /
• pp.153-169
• /
• 2015

The combined approach of using ego-centric network analysis and dynamic citation network analysis for refining the result of LDA-based topic modeling was suggested and examined in this study. Tow datasets were constructed by collecting Web of Science bibliographic records of White LED and topic modeling was performed by setting a different number of topics on each dataset. The multi-assigned top keywords of each topic were re-assigned to one specific topic by applying an ego-centric network analysis algorithm. It was found that the topical cohesion of the result of topic modeling with the number of topic corresponding to the lowest value of perplexity to the dataset extracted by SPLC network analysis was the strongest with the best values of internal clustering evaluation indices. Furthermore, it demonstrates the possibility of developing the suggested approach as a method of multi-faceted research trend detection.

• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.287-292
• /
• 2021

좋은 자연어 이해 시스템은 인간과 같이 텍스트에서 단순히 단어나 문장의 형태를 인식하는 것 뿐만 아니라 실제로 그 글이 의미하는 바를 정확하게 추론할 수 있어야 한다. 이 논문에서 우리는 뉴스 헤드라인으로 뉴스의 토픽을 분류하는 open benchmark인 KLUE(Korean Language Understanding Evaluation)에 대하여 기존에 비교 실험이 진행되지 않은 시중에 공개된 다양한 한국어 라지스케일 모델들의 성능을 비교하고 결과에 대한 원인을 실증적으로 분석하려고 한다. KoBERT, KoBART, KoELECTRA, 그리고 KcELECTRA 총 네가지 베이스라인 모델들을 주어진 뉴스 헤드라인을 일곱가지 클래스로 분류하는 KLUE-TC benchmark에 대해 실험한 결과 KoBERT가 86.7 accuracy로 가장 좋은 성능을 보여주었다.

• Journal of KIISE
• /
• v.43 no.11
• /
• pp.1201-1209
• /
• 2016

As the use of SNS continues to increase, various related studies have been conducted. According to the effectiveness of the topic model for existing theme extraction, a huge number of related research studies on topic model based analysis have been introduced. In this research, we suggested an automation system to analyze topics of each country and its distribution in twitter by combining world map visualization and issue matching method. The core system components are the following three modules; 1) collection of tweets and classification by nation, 2) extraction of topics and distribution by country based on topic model algorithm, and 3) visualization of topics and distribution based on Google geochart. In experiments with USA and UK, we could find issues of the two nations and how they changed. Based on these results, we could analyze the differences of each nation's position on ISIS problem.

• Journal of the Korean Institute of Landscape Architecture
• /
• v.51 no.3
• /
• pp.95-106
• /
• 2023

Korean healing research has developed over the past 20 years along with the growing social interest in healing. The field of healing research is diverse and includes legislated natural-based healing. In this study, abstracts of 2,202 academic journals, master's, and doctoral dissertations published in KCI and RISS were collected and analyzed. As for the research method, LDA topic modeling used to classify research topics, and time-series publication trends were examined. As a result of the study, it identified that the topic of Korean healing research was connected with 5 types and 4 mediators. The five were "Healing Tourism," "Mind and Art Healing," "Forest Therapy," "Healing Space," and "Youth Restoration and Healing," and the four mediators were "Forest," "Nature," "Culture", and "Education". In addition, only legalized healing studies extracted from Korean healing research and the topics were analyzed. As a result, legalized healing research classified into four. The four types were "Healing Spatial Environment Plan", "Healing Therapy Experiment", "Agricultural Education Experiential Healing", and "Healing Tourism Factor". Forest Therapy, which has the largest amount of research in legalized healing, Agro Healing and Garden Healing which operate similar programs through plants, and Marine Healing using marine resources also analyzed. As a result, topics that show the unique characteristics of individual healing studies and topics that are considered universal in all healing studies derived. This study is significant in that it identified the overall trend of research on Korean healing facilities and programs by utilizing LDA topic modeling.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2015.01a
• /
• pp.51-54
• /
• 2015

추천 시스템의 대표적인 연구 중 하나인 콘텐츠 기반 추천 시스템 연구는 TV 프로그램이나 영화의 줄거리, 장르, 리뷰 등의 콘텐츠의 메타데이터를 이용한다. 그러나 이러한 연구들은 콘텐츠 관련 정보에만 의존할 뿐, 시청자의 프로파일과 콘텐츠의 정보를 함께 고려하지 않는다. 본 논문에서는 시청자의 프로파일 중 연령과 콘텐츠의 정보인 프로그램의 줄거리를 활용한 TV 프로그램 추천 시스템을 제안한다. 본 추천 시스템은 시청자를 연령에 따라 분류한 후, LDA 알고리즘을 이용하여 시청자의 시청 TV 프로그램의 줄거리를 분류된 나이에 따라 각각의 줄거리 토픽 모델로 생성한다. 이를 기준으로 시청자가 원하는 시간대에 방송되는 프로그램들의 줄거리 토픽벡터와 시청자의 선호도 토픽벡터의 유사도를 비교해 가장 유사도가 높은 TV 프로그램을 시청자에게 추천하는 방식이다. 본 논문에서는 연구의 효용성을 검증하기 위해 줄거리만을 사용한 경우와 줄거리와 연령을 동시에 활용한 경우를 비교 실험하였다. 실험을 통해 프로그램의 줄거리만을 사용한 경우보다 연령을 동시에 활용한 경우의 추천 시스템 성능이 개선된 것을 확인할 수 있었다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.1
• /
• pp.88-95
• /
• 2021

This study aims to provide implications for establishing support policies for students by empirically analyzing data on university students dropouts. To this end, data of students enrolled in D University after 2017 were sampled and collected. The collected data was analyzed using topic modeling(LDA: Latent Dirichlet Allocation) technique, which is a probabilistic model based on text mining. As a result of the study, it was found that topics that were characteristic of dropout students were found, and the classification performance between groups through topics was also excellent. Based on these results, a specific educational support system was proposed to prevent dropout of university students. This study is meaningful in that it shows the use of text mining techniques in the education field and suggests an education policy based on data analysis.

• Annual Conference on Human and Language Technology
• /
• 2013.10a
• /
• pp.64-69
• /
• 2013

스마트폰의 대중화로 인하여 앱 마켓 시장이 급속도로 성장하였다. 이로 인하여 하루에도 수십개의 새로운 앱들이 출시되고 있다. 이러한 앱 마켓 시장의 급격한 성장으로 인해 사용자들은 자신이 흥미를 가질만한 앱들을 선택하는데 큰 어려움을 겪고 있어 앱 추천 방법에 대한 연구에 많은 관심이 집중되고 있다. 기존 연구에서 협력 필터링 기반의 추천 방법들을 제안하였으나 이는 콜드 스타트 문제를 지니고 있다. 이와는 달리 컨텐츠 기반 필터링 방식은 콜드 스타트 문제를 효율적으로 해소할 수 있는 방법이지만 앱설명에는 광고, 공지사항등 실질적으로 앱의 특징과는 무관한 노이즈들이 다수 존재하고 이들은 앱 사이의 유사관계를 파악하는데 방해가 된다. 본 논문에서는 이런 문제를 해결하기 위하여 앱 설명에서 노이즈에 해당하는 설명들을 자동으로 제거할 수 있는 모델을 제안한다. 제안하는 모델은 모바일 앱 설명을 구성하고 있는 각 문단을 LDA로 학습된 토픽들의 비율로 나타내고 이들을 분류문제에서 우수한 성능을 보이는 SVM을 이용하여 분류한다. 실험 결과에 따르면 본 논문에서 제안한 방법은 기존에 문서 분류에 많이 사용되는 Bag-of-Word 표현법에 기반한 문서 표현 방식보다 더 나은 분류 성능을 보였다.

• Journal of the Korean Society for Library and Information Science
• /
• v.40 no.2
• /
• pp.103-123
• /
• 2006

Assuming that the knowledge creation and retrieval functions could be the most important factors for a successful knowledge management(KM) especially during the promotion stage of KM, this study suggests an e-learning application as one of best methods for producing knowledge and also the integrated knowledge repository model in which learning, knowledge. and information objects can be semantically associated through topic map-based knowledge map. The traditional KM system provides a simple directory-based knowledge map. which can not provide the semantic links between topics or objects. The proposed model can be utilized as a solution to solve the above-mentioned disadvantages of the traditional models. In order to collect the basic data for the proposed model, first, case studies utilizing interviews and surveys were conducted targeting at three Korean insurance companies' knowledge managers(or e-learning managers) and librarians. Second, the related studies and other topic map-based pilot systems were investigated.

• The Korean Journal of Applied Statistics
• /
• v.34 no.5
• /
• pp.795-806
• /
• 2021

Recently, as text data through online channels have become vast, there is a growing interest in research that summarizes and analyzes them. One of the fundamental analyses of text data is to extract potential topics. Although the researcher may read all the data and summarize the contents one by one, it is not easy to deal with large amounts of data. Blei and Lafferty (2007) and Blei et al. (2003) proposed topic modeling methods for extracting topics using a statistical model. Since the text data is generally collected over time, it is worthwhile to monitor the topic's changes. In this study, we propose a topic index based on the results of the topic model. In addition, a control chart, a representative tool for statistical process management, is applied to monitor the topic index over time. As a practical example, we use text data collected from Blue House National Petition boards between March 5, 2018, and March 5, 2020.

• Information Systems Review
• /
• v.19 no.2
• /
• pp.1-19
• /
• 2017

Original scripts of e-learning lectures for the CEOs of corporation S were analyzed using topic analysis, which is a text mining method. Twenty-two topics were extracted based on the keywords chosen from five-year records that ranged from 2011 to 2015. Research analysis was then conducted on various issues. Promising topics were selected through evaluation and element analysis of the members of each topic. In management and economics, members demonstrated high satisfaction and interest toward topics in marketing strategy, human resource management, and communication. Philosophy, history of war, and history demonstrated high interest and satisfaction in the field of humanities, whereas mind health showed high interest and satisfaction in the field of in lifestyle. Studies were also conducted to identify topics on the proportion of content, but these studies failed to increase member satisfaction. In the field of IT, educational content responds sensitively to change of the times, but it may not increase the interest and satisfaction of members. The present study found that content production for CEOs should draw out deep implications for value innovation through technology application instead of simply ending the technical aspect of information delivery. Previous studies classified contents superficially based on the name of content program when analyzing the status of content operation. However, text mining can derive deep content and subject classification based on the contents of unstructured data script. This approach can examine current shortages and necessary fields if the service contents of the themes are displayed by year. This study was based on data obtained from influential e-learning companies in Korea. Obtaining practical results was difficult because data were not acquired from portal sites or social networking service. The content of e-learning trends of CEOs were analyzed. Data analysis was also conducted on the intellectual interests of CEOs in each field.