• Title/Summary/Keyword: 텍스트 연구

Search Result 3,494, Processing Time 0.031 seconds

A Study on Automated HoMokDan Structure Determination in Table (테이블 내의 호목단 구조 판별 자동화에 대한 연구)

  • Cho, Sung-Soo;Kim, Myung Ho
    • Annual Conference of KIPS
    • /
    • 2012.04a
    • /
    • pp.295-297
    • /
    • 2012
  • 현재 법률과 관련된 문서들은 변경 사항 에 대한 공표와 기록의 중요성을 가지고 있다. 따라서 변경사항을 자동으로 인지하고 공표할 수 있는 자동화 시스템에 대한 관심과 연구가 진행되고 있다. 그러나 대부분의 문서들은 복잡한 구조이기 때문에 자동화에 어려움이 많다. 이로 인해 복잡한 구조의 문서를 자동으로 판별할 수 있는 방법에 관한 관심이 증대되고 있다. 현재 국내외에서는 전자 문서 파일의 텍스트 및 테이블을 판별해서 분류 하는 자동화에 대한 연구가 진행되고 있다. 하지만 이전 연구에서는 호목단 구조를 갖는 계층적인 테이블을 판별하지 않는다. 그래서 본 논문에서는 호목단을 정의하고, 테이블의 호목단 구조를 패턴 별로 분류 하며, 테이블의 호목단 구조 판별 방법을 제시한다.

Implementation of an emotional subtitle editor for deaf and hearing impaired people (청각장애인을 위한 감성자막 편집기 구현)

  • Kim, Hyunsoon;Oh, Juhyun
    • Proceedings of the Korean Society of Broadcast Engineers Conference
    • /
    • fall
    • /
    • pp.120-123
    • /
    • 2021
  • 디지털화와 기술의 급격한 발전으로 방송 서비스도 고품질 서비스를 보다 편리하게 이용할 수 있도록 진화하고 있다. 이러한 변화하는 방송 환경에서 비장애인 대비 소외계층의 정보 접근성을 높이기 위한 연구에 대한 필요성이 증가하고 있다. 이러한 연구의 일환으로 UHD 자막 방송 서비스를 개선하기 위한 연구인 '감성표현 자막 서비스 기술' 연구를 진행하였다. 감성표현 자막 서비스 기술은 단순한 텍스트의 전달이 아닌 이미지와 폰트 스타일을 포함한 다양한 시각적 표현을 통해 청각장애인의 방송 내용에 대한 이해도를 향상시키기 위한 기술이다. 본 논문에서는 이러한 감성표현 자막 서비스를 소개하고 해당 서비스를 가능하게 하는 관련 기술과 시스템 구현 결과에 대하여 다룬다. 지상파 UHD 방송을 대상으로 개선된 형태의 자막 서비스를 제공하기 위한 핵심 시스템인 감성자막 편집기를 개발하였다. 감성자막 편집기는 화자의 감정 정보 등을 입력, 편집하고 편집된 감성자막을 영상과 싱크를 맞추어 재생하는 기술과 감성자막을 UHD 송출시스템으로 전송하는 시스템이다.

  • PDF

Construction of bilingually pre-trained language model from large-scaled Korean and English corpus (KE-T5: 한국어-영어 대용량 텍스트를 활용한 이중언어 사전학습기반 대형 언어모델 구축)

  • Shin, Saim;Kim, San;Seo, Hyeon-Tae
    • Annual Conference on Human and Language Technology
    • /
    • 2021.10a
    • /
    • pp.419-422
    • /
    • 2021
  • 본 논문은 한국어와 영어 코퍼스 93GB를 활용하여 구축한 대형 사전학습기반 언어모델인 KE-T5를 소개한다. KE-T5는 한국어와 영어 어휘 64,000개를 포함하는 대규모의 언어모델로 다양한 한국어처리와 한국어와 영어를 모두 포함하는 번역 등의 복합언어 태스크에서도 높은 성능을 기대할 수 있다. KE-T5의 활용은 대규모의 언어모델을 기반으로 영어 수준의 복잡한 언어처리 태스크에 대한 연구들을 본격적으로 시작할 수 있는 기반을 마련하였다.

  • PDF

Knowledge Graph Embedding Methods for Political Stance Prediction: Performance Evaluation (뉴스 기사의 정치적 성향 판단을 위한 지식 그래프 임베딩 기법의 효과 분석)

  • Seongeun Ryu;Yunyong Ko;Sang-Wook Kim
    • Annual Conference of KIPS
    • /
    • 2023.05a
    • /
    • pp.519-521
    • /
    • 2023
  • 온라인 뉴스 플랫폼의 발전은 에코 챔버(echo chamber) 효과와 정치적 양극화를 심화시키며, 이를 완화하기 위한 선행 연구로 뉴스 기사의 정치적 성향을 판단하는 연구가 필요하다. 기존 연구는 외부 지식 그래프를 활용하여 뉴스 기사의 텍스트 정보를 더욱 풍부하게 표현한다. 그러나, 외부 지식을 임베딩하는 지식 그래프 임베딩(knowledge graph embedding, KGE) 방법은 다양하며, 각 KGE 방법이 정치적 성향 예측 정확도에 미치는 효과에 대해서 충분히 연구되지 않았다. 본 논문에서는 정치적 성향 예측에 외부 지식의 활용을 최대화하기 위한 다양한 KGE 방법들의 효과를 분석한다. 실험 결과, 외부 지식 그래프 내의 개체들 간 복잡한 관계를 간단하고 정확하게 표현 가능한 ModE 방법을 활용하는 것이 정치적 성향 예측에 가장 효과적이라는 것을 확인하였다.

Long-KE-T5: Korean-English Language model for Long Sequences (Long-KE-T5: 긴 맥락 파악이 가능한 한국어-영어 언어 모델 구축)

  • San Kim;Jinyea Jang;Minyoung Jeung;Saim Shin
    • Annual Conference on Human and Language Technology
    • /
    • 2023.10a
    • /
    • pp.168-170
    • /
    • 2023
  • 이 논문에서는 7,400만개의 한국어, 영어 문서를 활용하여 최대 4,096개의 토큰을 입력으로하고 최대 1,024개의 토큰을 생성할 수 있도록 학습한 언어모델인 Long-KE-T5를 소개한다. Long-KE-T5는 문서에서 대표성이 높은 문장을 생성하도록 학습되었으며, 학습에 사용한 문서의 길이가 길기 때문에 긴 문맥이 필요한 태스크에 활용할 수 있다. Long-KE-T5는 다양한 한국어 벤치마크에서 높은 성능을 보였으며, 사전학습 모델링 방법이 텍스트 요약과 유사하기 때문에 문서 요약 태스크에서 기존 모델 대비 높은 성능을 보였다.

  • PDF

The Research Trend Analysis of the Korean Journal of Physical Education using Mecab-ko Morphology Analyzer (Mecab-ko 형태소 분석을 이용한 한국체육학회지 연구동향 분석)

  • Park, Sung-Geon;Kim, Wanseop;Lee, Dae-Taek
    • 한국체육학회지인문사회과학편
    • /
    • v.56 no.6
    • /
    • pp.595-605
    • /
    • 2017
  • The purpose of this study is to investigate what kind of research fields are preferred by the researcher of the Korean Physical Education Society using the Mecab-ko morpheme analysis and whether there are differences in the interests of researchers between the humanities and social sciences and natural sciences. A total of the data collected for this study are 5,014 papers published online from March 2002 to March 2017 in the Korean Journal of Physical Education was collected. In this study, we used Mecab-ko morpheme analyzer to extract the keyword from the collected documents. As a result, the study found that the number of papers published in KAHPERD appeared to be decreasing. It was also that the main concern of researchers in KAHPERD toward was leisure, live sports and health were relatively higher than the improvement of performance. The research subjects that were interested in the research were women, middle-aged and elderly. The study found that researchers in the humanities and social sciences have shown interest in both traditional research and social interests, while researchers in the natural sciences have shown an interest in a deeper study of traditional research. In conclusion, in order to realize the revitalization of sports convergence research, it is necessary to establish standards for the field of study which should focus on the depth and breadth of research.

Analysis of the AI Convergence Science Education Research Trends Using Text Mining (텍스트 마이닝을 활용한 AI융합 과학교육 연구 동향 분석)

  • Lee, Ju-Young
    • Journal of Korean Elementary Science Education
    • /
    • v.43 no.4
    • /
    • pp.544-553
    • /
    • 2024
  • The purpose of this study was to analyze the trends of research focusing on artificial intelligence and the science education and derive important problems, topics, and research trends,. The analysis of the AI convergence science education research trends targeted 83 articles on the awareness of artificial intelligence, research trends, design, development, and application of the education programs related to artificial intelligence. The analysis data was collected through the RISS. The collected data was refined using Excel and Textom, and the main keywords were identified and analyzed through the frequency analysis and keyword network analysis. The connection centrality of the keywords was confirmed using the CONCOR analysis. The research results showed that the AI convergence science education research was expanding in both quantitative and qualitative aspects, and that the main keywords were identified as 'AI,' 'AI convergence education,' 'AI convergence science education,' 'AI education,' 'science education,' 'science,' 'machine learning,' 'elementary school,' 'generative AI,' and 'educational program.' Through the connection centrality analysis and CONCOR analysis, it was confirmed that the clusters were formed around the 'naming,' 'content and method,' 'elementary,' and 'data' in the AI integrated science education. Based on the results, the main topics and trends of the research integrating artificial intelligence into the science subjects were derived and the implications and directions for follow-up research were set forth.

Search for an archaic form of Jain-Danoje - Focucing on 'Yeowonmoo' and 'Hojanggut' - (자인단오제의 고형(古形)에 관한 탐색 - '여원무'와 '호장굿'을 중심으로 -)

  • Han, Yang-myung
    • (The) Research of the performance art and culture
    • /
    • no.19
    • /
    • pp.5-33
    • /
    • 2009
  • Jain-Danoje's course since modern is not different with almost all of folk performances, which were restored and reconstructed with a background of the designation of an intangible cultural heritage and National folk arts contest sine the 1960s. Generally, these folk performances were decontextualized in course of extinction and reappearance, and recontextualized in course of new directions on tradition. Also, the performances were interpreted differently and transformed by the main constituents of reappearance. Jain-Danoje nowadays has a regular form just at that time that has been designated as a cultural heritage at 1970s. But, today's Jain-Danoje is clearly different with the last appearance in 1936 and some Literature and jainhyun-eupji. I think such differences would stems from the process of reproduction. From this perspective, I had investigate Old literature and the early days report, and the current text. Especially, I will show the considerable change which has been occurred in the Yeowonmu and Hojanggut, the central role to configure that identity, by comparing past and today. As a result of consideration, today's form of the Yeowonmu and Hojanggut are created texts that mind the designation of an intangible cultural heritage and National folk arts contest. These texts has been reproduced without understanding about structure and current of folk festival and state of performance which has been transmitted on premodern society. some intellectuals search for an archaic form of Jain-Danoje based on jainhyun-eupji that created in 1895, except the other jainhyun-eupji. Moreover, because of the understanding with a bias, they can't grasp the meaning about the religious service for Hanjanggun, and they can't see the facts of Yeowonmoo. In addition, they were aware of 'o-sin' that led by Hojang as a fancy dress parade in a carnival, and that is recognized as a component of Jain-Danoje, so there was other text which is different from our own festival.

The Aesthetics of Conviction in Novel and Film Mephisto (소설과 영화 속 '메피스토'의 사상성 미학)

  • Shin, Sa-Bin
    • Journal of Popular Narrative
    • /
    • v.25 no.1
    • /
    • pp.217-247
    • /
    • 2019
  • This research paper intends to examine the intertextuality of Klaus Mann's novel Mephisto (1936) and István Szabó's film Mephisto (1981) and how the derivative contents (i.e., film) accepted and improved the schematic aesthetics of conviction in original contents (i.e., novel). In general, the aesthetics of conviction is applied to criticize the state socialism of the artists of the Third Reich or the ideology of the artists of East Germany from a biased ethical perspective. Mephisto is also based on the aesthetics of conviction. Thus, it would be meaningful to examine the characteristic similarity and difference between Klaus Mann's real antagonist (i.e., Gustaf Gründgens) and fictional antagonist (i.e., Hendrik Höfgen) from a historical critical perspective. In this process, an aesthetic distance between the real and fictional antagonists would be secured through the internal criticism in terms of intertextuality. In this respect, the film aesthetics of István Szabó are deemed to overcome the schematic limit of the original novel. The conviction in both the novel and film of Mephisto pertains to the belief and stance of a person who compromised with the state socialism of Nazi Germany, i.e., succumbed to the irresistible history. Klaus Mann denounced Mephisto's character Höfgen (i.e., Gründgens in reality) as an "Mephisto with evil spirits" from the perspective of exile literature. For such denunciation, Klaus Mann used various means such as satire, caricature, sarcasm, parody and irony. However, his novel is devoid of introspection and "utopianism", and thus could be considered to allow personal rights to be disregarded by the freedom of art. On the contrary, István Szabó employed the two different types of evil (evil of Mephisto and evil of Faust) from a dualistic perspective (instead of a dichotomous perspective of good and evil) by expressing the character of Höfgen like both Mephisto and Hamlet (i.e., "Faust with both good and evil spirits). However, Szabó did not present the mixed character of "Mephisto and Hamlet (Faust)" only as an object of pity. Rather, Szabó called for social responsibility by showing a much more tragic end. As such, the novel Mephisto is more like the biography of an individual, and the film Mephisto is more like the biography of a generation. The aesthetics of conviction of Mephisto appears to overcome biased historical and textual perspectives through the irony of intertextuality between the novel and the film. Even if history is an irresistible "fate" to an individual, human dignity cannot be denied because it is the "value of life". The issue of conviction is not only limited to the times of Nazi Germany. It can also be raised with the ideology of the modern and contemporary history of Korea. History is so deeply rooted that it should not be criticized merely from a dichotomous perspective. When it comes to the relationship between history and individual life, a neutral point of view is required. Hopefully, this research paper will provide readers with a significant opportunity for finding out their "inner Mephisto" and "inner Hamlet."

The Research Trends in Journal of the Korean Institute of Landscape Architecture using Topic Modeling and Network Analysis (토픽모델링과 연결망 분석을 활용한 국내 조경 분야 연구 동향 분석 - 한국조경학회지를 대상으로 -)

  • Park, Jae-Min;Kim, Yong Hwan;Sung, Jong-Sang;Lee, Sang-Seok
    • Journal of the Korean Institute of Landscape Architecture
    • /
    • v.49 no.2
    • /
    • pp.17-26
    • /
    • 2021
  • For the past half century, the Journal of the Korean Landscape Architecture has been leading the landscape architecture research and industry inclusively. In this study, abstracts of 1,802 articles were collected and analyzed with topic modeling and network analysis method. As a result of this paper, a total of 27 types of subjects were identified. Health and healing in the field of environmental psychology, garden and aesthetics, participation and community, modernity, place and placenness, microclimate, tourism and social equity also have been continued as important research area in this journal. Modernity, community and urban regeneration is hot topics and ecological landscape related topics were cold topics. Although there was a difference by subject, the variability of the research subjects appeared after the 2000s. In Network analysis, it shows that 'Park' is a representative keyword that can symbolize the journal, and 'landscape' is also important a leading area of the journal. Looking at the overall structure of the network, it can be seen that the journal conducts research on 'utilizing', 'using', and creating 'park', 'landscape', and 'space'. This study is meaningful in that it grasped the overall research trend of the journal by using topic modeling and network analysis of text mining.