국어교육을 위한 텍스트 분류 체계 연구 -읽기와 쓰기 영역을 중심으로-

김봉순 ( Bong Soon Kim ) 국어교육학회 2010 國語敎育學硏究 Vol.39 No.-

이 연구는 국어교육에 적합한 텍스트 분류 체계를 개발하여 제시하고자 하였다. 이를 위하여 수사학, 문예학, 텍스트언어학에서 논해진 기존의 분류법을 국어교육의 관점에서 비판적으로 검토하여 국어교육에 적절한 분류 기준을 설정하고, 이 기준에 따른 분류 체계를 예시하였다. 기존 연구를 국어교육의 관점에서 검토한 결과, 첫째, 기존의 ``기능 중심 분류법`` 대신 ``구조 중심 분류법``이 유용하며, 둘째, 구조보다 더 상위의 개념으로 ``자아``와 ``세계``의 ``내용 성격``의 범주를 설정하고, 셋째, 구조 범주의 하위에 ``장르 관습``의 범주를 두는 것이 필요하다고 보았다. 이에 따라 ``내용 성격, 결속 양식, 장르 관습``의 세 기준을 차례로 적용하여 텍스트를 분류하였다. 그 결과, 첫째, 텍스트는 ``자아 표명형``, ``세계설명형``, ``세계·자아상응형``의 세 가지로 크게 분류된다. 둘째, 이 세 유형은 각각 ``주장-근거``형 등, ``설명형`` 등, ``사실-생각형`` 등으로 분류된다. 셋째, 이들 하위 유형은 다시 ``논설문류, 선언문류, 논평류 등``의 구체적인 장르로 분류된다. 이와 같이 설정된 텍스트분류 체계는, 국어과 교육과정과 교과서를 구성할 때와 읽기 및 쓰기 영역에서 글 특성에 관한 지식을 가르칠 때에 효과적인 지식기반이 될 것으로 기대된다. The purpose of this study is to develop and suggest the classification of texts that is fit for teaching reading and writing in Korean language education. For this purpose, I criticized the prior researches on a classification of texts in the areas of rhetorics, literary criticisms, and textlinguistics. And I abstracted the three categories to classify texts for teaching reading and writing. As the results of the survey and critiques prior researches, I concluded that text-structures were more important and fitter than text-functions when we classified texts for Korean language education. And I added ``contents: what is said in texts`` and ``conventions: what are conventional forms of texts`` to make the classification more concrete. So, I classified texts by the new three category: contents, text-structures, and conventions. With the first category ``contents``, I classified thee types of texts: self-expression types, object-explanation types, and object and self-correspondence types``. With the second category ``text-structures``, I subclassified each type of texts. And with the third category ``conventions``, I exemplified some kinds of genres.

작문 ‘텍스트 분류’ 방안 연구

김종률(Kim, Jong-ryul) 한국어문학회 2012 語文學 Vol.0 No.115

The purpose of this study is to look at the problems of text classification described in the revised Korean language curriculum in 2009 and to set-up systematization of text classification for problem-solving around diachronic side of essay. Text classification of the revised Korean language curriculum in 2009 is divided into text for information delivery, text for persuasion, text for social interaction, text for self-reflection, text for learning. Among them, the text for learning is excluded in the revised Korean language curriculum in 2011. But the problem of the relationship set with essay writing belonging to this text type and editorial writing, proposal writing, critique, comments on current events etc. belonging on text for persuasion still has potential. To solve this problem, we will have to consider diachronic side of essay. the term, essay was widely used in personal anthology of Goryeo and Joseon Dynasty. When essay is related to writing form, it is associated with non-writing[論文], seol-writing[說文], won-writing[原文], byeon-writing[辯文], pyeong-writing[評文], ui-writing[議文], hae-writing[解文], sangso-writing[上疏文], chaek-writing[策文] etc. The style of traditional essay continues that tradition through a writing formal union of the modern sense, not gone. In other words, non-writing and seol-writing changes as editorial writing, pyeong-writing turns into critique and comments on current events, ui-writing and sangso-writing changes as proposal writing. Therefore the text for persuasion can be set up into essay style in text classification of writing education. And the text for information delivery can be set up into description style, the text for social interaction and self-reflection can be set up into narrative style. This new text classification could be one way on the existing text classification.

CNN-based Skip-Gram Method for Improving Classification Accuracy of Chinese Text

( Wenhua Xu ),( Hao Huang ),( Jie Zhang ),( Hao Gu ),( Jie Yang ),( Guan Gui ) 한국인터넷정보학회 2019 KSII Transactions on Internet and Information Syst Vol.13 No.12

Text classification is one of the fundamental techniques in natural language processing. Numerous studies are based on text classification, such as news subject classification, question answering system classification, and movie review classification. Traditional text classification methods are used to extract features and then classify them. However, traditional methods are too complex to operate, and their accuracy is not sufficiently high. Recently, convolutional neural network (CNN) based one-hot method has been proposed in text classification to solve this problem. In this paper, we propose an improved method using CNN based skip-gram method for Chinese text classification and it conducts in Sogou news corpus. Experimental results indicate that CNN with the skip-gram model performs more efficiently than CNN-based one-hot method.

문장 레벨 그래프 회선 신경망을 통한 텍스트 분류

이민우,김양훈,정교민 한국정보과학회 2019 정보과학회 컴퓨팅의 실제 논문지 Vol.25 No.8

텍스트 분류는 자연어처리 분야의 전통적인 문제이다. 기존의 RNN 및 CNN 기반 텍스트 분류 모델들은 순차적인 단어 구조에 의존하기 때문에 인접하지 않지만 관련성이 높은 단어 간의 관계를 유추하기 어렵다는 문제점이 있다. 반면 GCN(Graph Convolutional Network)은 그래프의 형태로 데이터를 입력받기 때문에 문장의 순차적 구조에 대한 의존도를 줄일 수 있다. 본 논문에서는 문서의 비순차적인 관계를 그래프로 담아내어 더욱 효과적으로 파악하고 분류하는 인공신경망 모델을 제안한다. 문서를 그래프로 표현하기 위해 각 단어를 그래프의 노드로 변환하고, 단어 간의 관계를 계산해 엣지로 정의한다. 최근에 제시된 GCN 구조를 통해 단어 간의 관계가 반영된 단어 벡터를 계산한 뒤, 어텐션 기반 요약 함수를 통해 문단을 주어진 클래스로 분류하는 방법을 제시한다. 실험 결과, 새롭게 제시된 모델이 RNN 및 CNN 기반 텍스트 분류 모델보다 좋은 성능을 보였다. Text classification is an important task in natural language processing, and most of the recent approaches employ neural networks to learn and classify the texts. RNN and CNN based models, which are widely used for solving the task, involve reading and processing the text in a sequential manner. This creates inefficiency in learning dependencies between far-apart words. On the contrary, Graph Convolutional Network (GCN) architecture is capable of processing more complex graph-structured data, thus having potential to recognize and learn from complex linguistic structures. In the present work, we transform text sequences into graphs by assigning each word in the text as a node and representing the relationship between words as edges. We then propose a method for solving text classification that uses recent GCN architectures to take the transformed text-graph as input, learn hidden representations, and output a single hidden representation for classification. In our experiments, our proposed model outperformed RNN and CNN based models with regards to various text classification tasks.

Issues and Empirical Results for Improving Text Classification

Youngjoong Ko,Jungyun Seo 한국정보과학회 2011 Journal of Computing Science and Engineering Vol.5 No.2

Automatic text classification has a long history and many studies have been conducted in this field. In particular, many machine learning algorithms and information retrieval techniques have been applied to text classification tasks. Even though much technical progress has been made in text classification, there is still room for improvement in text classification. In this paper, we will discuss remaining issues in improving text classification. In this paper, three improvement issues are presented including automatic training data generation, noisy data treatment and term weighting and indexing, and four actual studies and their empirical results for those issues are introduced. First, the semi-supervised learning technique is applied to text classification to efficiently create training data. For effective noisy data treatment, a noisy data reduction method and a robust text classifier from noisy data are developed as a solution. Finally, the term weighting and indexing technique is revised by reflecting the importance of sentences into term weight calculation using summarization techniques.

Issues and Empirical Results for Improving Text Classification

Ko, Young-Joong,Seo, Jung-Yun Korean Institute of Information Scientists and Eng 2011 Journal of Computing Science and Engineering Vol.5 No.2

Automatic text classification has a long history and many studies have been conducted in this field. In particular, many machine learning algorithms and information retrieval techniques have been applied to text classification tasks. Even though much technical progress has been made in text classification, there is still room for improvement in text classification. In this paper, we will discuss remaining issues in improving text classification. In this paper, three improvement issues are presented including automatic training data generation, noisy data treatment and term weighting and indexing, and four actual studies and their empirical results for those issues are introduced. First, the semi-supervised learning technique is applied to text classification to efficiently create training data. For effective noisy data treatment, a noisy data reduction method and a robust text classifier from noisy data are developed as a solution. Finally, the term weighting and indexing technique is revised by reflecting the importance of sentences into term weight calculation using summarization techniques.

The SVM based Uyghur Text Classification and its Performance Analysis

Palidan Tuerxun,Fang Dingyi,Askar Hamdulla 보안공학연구지원센터 2015 International Journal of Multimedia and Ubiquitous Vol.10 No.4

This paper mainly explores the use of Support Vector Machines (SVMs) for Uyghur text classification, presents the process of text categorization: Text preprocessing, feature dimensionality reduction, representation method and classification of text features etc., discusses the SVMs classification algorithm in the application of Uyghur text classification. Focus on the construction of text categorization model and its procedures. Experiment results show that training by using the selected training data with the guarantee of the performance of the classifier, has higher efficiency than other nearest neighbor classifier (KNN), Naive Bayes (NB) classifier with increased accuracy.

한국어 읽기 교육을 위한 텍스트 유형 연구

김재희(Kim, Jae-hee),안정오(An, Cheung-o) 고려대학교 한국학연구소 2015 한국학연구 Vol.54 No.-

본고는 한국어 읽기 텍스트의 분류 작업을 위한 기준을 제시하고, 향후에 현장에서 시험적으로 적용할 수 있는 기호학적 텍스트 유형의 읽기 교육을 위한 방안을 마련하는 데 목적이 있다. 그간에 한국어 교육에서 읽기 텍스트의 유형 분류에 관한 연구 결과가 활발하게 나왔지만 주로 텍스트의 기능, 장르, 구조 등을 분류 기준으로 삼고 교재를 분석했다. 기존의 분류법은 텍스트 간의 유사성을 밝혀내고 구성적 자질을 체계화한다는 점에서 의미가 있지만 분석틀이 너무 상세해서 개별 텍스트의 특징을 단편적이고 지엽적으로 분석한다는 한계를 가진다. 따라서 텍스트의 분류 기준을 명료화해서 텍스트에 본질적으로 접근할 필요가 있다. 그래서 본고는 퍼스의 기호학에 나오는 ‘도상, 지표, 상징’의 개념을 텍스트에 적용하여 각 텍스트의 특성을 정리해 보고자 한다. 퍼스의 도상, 지표, 상징이론에 근거한 분류는 거시적이고 포괄적인 분류 방법이며 저자와 독자가 작성된 텍스트와 작성될 대상 간의 관계성을 느낄 수 있는 분류법이라고 볼 수 있다. 또한 이 분류 체계는 텍스트의 난이도를 선정하는 데도 판단 기준을 제공할 수 있을 것이다. 이러한 퍼스의 기호학적 분류법으로 한국어 교재를 도상, 지표, 상징 순으로 배치한다면 인지적이고 논리적 수준에 따른 분류가 되며 새로운 한국어 교육 방식의 단초를 제공하는 시도가 될 것이다. This study aims to provide a standard for the classification of Korean reading texts and to present an effective plan for reading education. Many studies have classified texts, but most analysis was based on text function, genre, structure, etc. Existing classification methods are limited; they only pay secondary attention to characteristics of individual texts. Therefore, simplification is necessary for a more fundamental approach. In this study, the concepts of ‘icon, index and symbol’ from the semiotics of Charles S. Peirce are used to organize textual characteristics. The classification system in this study is macroscopic and comprehensive, and is critical to the thinking process for understanding the texts. Furthermore, it can provide a standard of judgment to determine the level of difficulty of texts. Namely, iconic texts describe the object itself so there is low cognitive load, while Index texts are analyzed by logical correlation, even though there is an existential connection, and Symbolic texts are understood by arbitrary agreement or convention; the latter two have a high cognitive load. If Peirce’s semiotic classification method can be applied to Korean textbooks, this would be a cognitive and logical level of classification and would mean a new approach for Korean language education.

한국어 표준 교육과정의 텍스트 영역 특성 연구

이은희 한국텍스트언어학회 2020 텍스트언어학 Vol.48 No.-

This study examines the characteristics of the text area of the Korean standard curriculum as a foreign language through the review of the internal system of the curriculum and its comparison with other curricula. Chapter 2 investigates how the text area is set up in the course of developing the curriculum, and chapter 3 compares and reviews Korean standard curriculum as a foreign language and Korean language-related curricula in which text is set as a component of the content system. The status of the text area in the Korean standard curriculum as a foreign language appears to change in the curriculum development process, but this is only a change in the normal position of the curriculum, and what is actually dealt with is limited to the aspect of educational materials. Looking at the way text area is organized in three ways, text items selection, text grade, and text classification, there are some problems in each part, but particularly with regard to text classification, it is necessary to review it in various directions such as terminology, characteristics of categories, and relationships among categories. Among Korean language- related curricula, the KSL curriculum appears simple, without the type classification of text. On the other hand, the reading curriculum of the Korean as a foreign language combines the text type and the structural knowledge of the text necessary to understand the text, while the Korean language curriculum presents text items as educational materials and content elements for text type education according to the characteristics of the language use area. In the case of the Korean standard curriculum as a foreign language, more diverse approaches are needed to educationally implement the functions of text in language education.

DCT와 정보 화소 밀도를 이용한 PDA로 획득한 명함 영상에서의 영역 해석

김종흔,장익훈,김남철 한국통신학회 2004 韓國通信學會論文誌 Vol.29 No.8C

본 논문에서는 PDA에 장착된 카메라를 사용하여 획득한 명함 영상에 대한 효율적인 영역 해석 알고리듬을 제안한다. 제안된 방법은 크게 영역 분할, 정보 영역 분류, 문자 영역 분류의 3개 과정으로 구성된다. 영역 분할에서는 입력 명함 영상을 8${\times}$8 크기의 블록으로 나누고 각 블록을 저주파 대역에서의 정규화 된 DCT 계수의 에너지를 이용하여 정보 블록과 배경 블록으로 분류한 다음, 블록에 대한 영역 라벨링을 통하여 정보 영역과 배경 영역으로 분할한다. 정보 영역 분류에서는 각 정보 영역을 블록 신호의 수평, 수직 방향 에지 성분과 저주파 대역에서의 DCT 계수의 에너지 비와 이진화 된 정보 영역 내에서의 흑화소인 정보 화소의 밀도를 이용하여 문자 영역과 배경 영역으로 분류한다. 문자 영역 분류에서는 분류된 문자 영역을 정보 화소의 밀도와 평균 런 길이를 이용하여 다시 큰 문자와 작은 문자 영역으로 분류한다. 실험결과 제안된 영역 해석 방법은 여러 종류의 명함을 다양한 주변 여건에서 PDA로 획득한 시험 영상에 대하여 정보 영역과 배경 영역을 잘 분할하고, 정보 영역을 문자 영역과 그림 영역으로 잘 분류하며, 다시 문자 영역을 큰 문자와 작은 문자 영역으로 잘 분류함을 보였다 그리고 제안된 영역 분할 방법과 정보 영역 분류 방법은 기존의 방법들보다 각각 약 2.2-10.1%와 7.7%의 에러율 향상을 보였다. In this paper, we present an efficient algorithm for region analysis of business card images acquired in a PDA by using DCT and information pixel density. The proposed method consists of three parts: region segmentation, information region classification, and text region classification. In the region segmentation, an input business card image is partitioned into 8 f8 blocks and the blocks are classified into information and background blocks using the normalized DCT energy in their low frequency bands. The input image is then segmented into information and background regions by region labeling on the classified blocks. In the information region classification, each information region is classified into picture region or text region by using a ratio of the DCT energy of horizontal and vertical edge components to that in low frequency band and a density of information pixels, that are black pixels in its binarized region. In the text region classification, each text region is classified into large character region or small character region by using the density of information pixels and an averaged horizontal and vertical run-lengths of information pixels. Experimental results show that the proposed method yields good performance of region segmentation, information region classification, and text region classification for test images of several types of business cards acquired by a PDA under various surrounding conditions. In addition, the error rates of the proposed region segmentation are about 2.2-10.1% lower than those of the conventional region segmentation methods. It is also shown that the error rates of the proposed information region classification is about 1.7% lower than that of the conventional information region classification method.