통사 기반 한국어 대명사 해석

조은경(Eun Kyoung Jo),한정한(Jeong Han Han),김진규 (Jin Gyu Kim) 한국HCI학회 2003 한국HCI학회 학술대회 Vol.2003 No.2

텍스트 말뭉치의 주제와 주제어 분석 -세종 문어 말뭉치를 중심으로-

조은경 ( Eun Kyoung Jo ),한영균 ( Gyun Young Han ) 한국텍스트언어학회 2014 텍스트언어학 Vol.37 No.-

This paper is about extracting topic words and Topic categories on text corpora. For this purpose, we utilize statistical method of topic modeling named ‘LDA(Latent Dirichlet Allocation)’ and word weighting scheme named TF-IDF(Term Frequency-Inversed Document Frequency). As corpus material, We used Sejong part-of-speech tagged corpus for this analysis. After extracting topic words from corpus automatically, we investigate the topics and the topic words in the Sejong Corpus with the predefined categories by Sejong corpus constructors and designers and then we show that a topic come out by several topic words co-occurred rather than by a single topic word itself. We are sure that text analytic methods used in this paper would be very useful afterwards since our world will have more electronic texts and we will be able to analyze more corpus statistically.

감성 분석 연구의 현황과 말뭉치에 기반한 사례 분석

조은경(Eun Kyoung Jo) 언어과학회 2012 언어과학연구 Vol.61 No.-

There are many researches on the sentiment analysis as the application of the semantic analysis in the area of computational linguistics and the application of linguistic informatics. We have reviewed the international researches and national researches on the approaches. And we have performed a case study based on the movie reputation corpus. We have observed the lexicon which takes a key role in the sentiment analysis by showing how the sentiment analysis is successfully performed in a corpus based on the sentimental lexicon extracted from the other experimental corpus.

세종전자사전 활용을 위한 자연어 생성적 접근

조은경(Jo Eun-kyoung) 한글학회 2016 한글 Vol.- No.311

세종전자사전(이하 세종사전)은 한 어휘 항목이 형태 정보, 의미 정보, 통사 정보를 가지고 여러 어휘들의 관계가 기입된 통합 어휘부이며 여러 하위 범주 사전으로 구성된 대규모 지식 베이스이다. 세종사전은 다양한 응용을 지향한 산출물임에도 불구하고 사전의 내용에 대한 연구와 자연어 처리에서의 단방향 접근에 치우쳐 있다. 무엇보다 그 산출물을 재가공하여 활용하려는 노력이 드물다. 이 논문은 자연어 생성의 방향에서 접근하여 전산 어휘부로서 세종사전의 미시 구조를 살피고, 복잡한 미시 구조의 개별 어휘 항목을 문장생성 과정에 선택적으로 활용하기 위한 격자 형식화를 제안한다. 격자 형식화는 사전의 복잡한 기술 내용을 단순한 구조로 표현되게 하는 것이다. 이는 개별 어휘 항목이 아니라 여러 어휘 항목의 정보를 보게하고, 어휘부 정보를 자연어로 쉽게 구성하게 한다. 특히, 용언과 체언사전의 격자 형식화는 표현 구조를 쉽게 짜게 한다. 그리고 어미 사전의 격자 형식화는 표현 구조에서 표층 형태로 변환을 쉽게 한다. 우리는 통합 지식 베이스로서 세종사전의 활용을 위한 자연어 생성적 접근으로서 세종사전의 미시 구조를 살피고, 가능 표현 구조와 표층적인 어휘 형태의 생성을 위한 격자 형식화를 설계하였다. 또 이러한 과정을 통해 전산 어휘부로서의 개선점을 정리하였다. Sejong Dictionary is a knowledgebase platform which consists of various kinds of subcategory dictionaries and also a huge synthetic lexicon with morphologica, semantic and syntatic information. Although Sejong dictionary was the product of the project which was oriented to various application uses, it has been approached by a few studies of NLP and there is scarcely an effort to manipulate and use it. This paper approaches to Sejong dictionary from NLG study and investigates the microstructure of it and then set out to format the lattice structure as a design of information makeup for developing NLG. The lattice formating is to make from the complex structure into the simple structure. This makes us have the information of various lexical entries instead of that of individaul lexical item and also makes it easy to construct natural language expression from lexicon. For one, we show that the representing structure of predicate and argument can be easily composed on lattice formatting. In addition, we show that the surface expressions also can be easily converted from the representing structures on it. Furthermore, in the process of this examination, we figure out the improvements in the macro and micro structure of this dictionary for a computational feasibility. In all, we approached to Sejong dictionary from NLG perspective. we have examined the micro structure and tried to design the lattice formatting for computationally generating. During these process we arrange the improvements for computational lexicon.

인간 기계 상호작용을 위한 대화 처리 연구

조은경(Jo Eun-kyoung) 한글학회 2014 한글 Vol.- No.306

이 논문은 여행 영역의 인간 고객-에이전트의 목적 지향 대화 말뭉치 분석을 통해 인간 기계간 상호작용을 위한 대화 처리 모형을 구한다. 첫 번째, 대화의 구조적 측면에 있어, 대화 말뭉치로부터 발화 인접쌍을 추출하여 인접쌍 모델이 대화 관리에 유용함을 보인다. 그리고 여행 에이전트의 고객에 대한 응답 유형을 조사함으로써 대화 목적에 도달하기 위해 기계가 어떻게 인간 에이전트같이 대응할 수 있는지를 제안한다. 두 번째, 발화 단위 측면에 있어 자동 분류 방식을 활용하여 자연 발화를 어떻게 자동으로 의미 표상할 수 있을지에 대해서 실험하고, 이러한 방법은 태깅 말뭉치의 확장에도 유용함을 보인다. 그리고 말뭉치 분석 과정에서 여행 예약에 관한 발화 단위 의미 표상을 위한 영역 지식을 구축하였다. We try to get dialog processing model for man-machine interaction through corpus analysis on goal-oriented dialog of human customer-agent in travel domain. First, We get extract utterance adjacency pairs from dialog corpus and show that the pair model is useful for dialog management. And, We investigated the pattern of travel agent’s response utterances to customer and suggest how machine can response to man as well as human agent for reaching dialog goal. Second, We experimented how semantic representation of man utterance can be made automatically and show it is also useful for increasing tagged corpus. And we tried to build domain knowledge for semantic representation of utterance about travel reservation while investigating dialog corpus.

공지시와 연관 어휘에 의한 텍스트 응집성 분석

조은경 ( Jo Eun Kyoung ) 서강대학교 언어정보연구소 2017 언어와 정보 사회 Vol.31 No.-

Text coherence can be addressed to be built by grammatical cohesion and lexical cohesion. Coreference between a pronoun and its antecedent is one of grammatical cohesion devices. Relatedness among lexical items establishes lexical cohesion in a discourse. Centering theory was keenly aware of coreference which shows centering transition based on grammatical role and makes a discourse coherent. It has been numerously utilized as a method for interpreting pronouns, but, not for coherence analysis. Lexical relatedness has been sought from external lexical resources such as thesaurus or lexicon. However, it can be achieved by distributional models of their occurrences in various subject article texts. Some important lexical items which contribute coherence in certain articles are not in the other subject articles and they go together at the same time in a subject article. This lexical information can be schemed out using term weighting and relative frequency. All in all, we demonstrate how to analyse textual coherence quantitatively by centering transition and lexical distribution.

대화 시스템을 위한 대화 분석과 맥락 활용

조은경 ( Eun Kyoung Jo ) 한국텍스트언어학회 2015 텍스트언어학 Vol.39 No.-

The goal of this paper is to do dialog analysis and get a model of using dialog context for task-oriented dialog system. Among the recent problems of dialog processing which many researchers are struggling with is to use dialog context for determing the speech act and interpreting the semantics of it. Because,it is difficult to figure out which information or features can be formalized for dialog context.To manage the dialog context, we would like to show that we should use speech act sequence longer than adjacency pair and the features about subdialog structure and dialog goal. For this, we made a dialog analysis of investigating dialog contexts based on dialog corpus which has been annotated with semantic tags for dialog processing. For one,we quantitatively examined the adjacency pairs and proposed the speech act sequence instead of the adjacency pair. In addition, we showed that subdialog structure could be considered as a speech act sequence pattern. Furthermore, we showed that specification features for dialog goal could be considered. Finally, we made an experiment on determine speech acts using dialog contexts which have been considered as the contextual features through corpus-based investigation. To take into account using the various features of dialog context for constructing the model of dialog system, we build maximum entropy modelling with generalized iterative scaling. Also,we showed that our experiment of using dialog context with speech act sequence could be accepted statistically as a fair one.

통사 기반 한국어 대명사 해석

조은경(Eun Kyoung Jo),한정한(Jeong Han Han),김진규(Jin Gyu Kim) 한국정보과학회 2003 한국정보과학회 학술발표논문집 Vol.2003 No.2

디지털 사회의 가치와 언어 연구 방법

조은경 ( Eun Kyoung Jo ) 서강대학교 언어정보연구소 2016 언어와 정보 사회 Vol.28 No.-

In this digital society, we seek the covenience even in using language, surrounded by digital environment. A lot of communication take place in online space, through digital computer. Therefore, we can research language based on the digital text resource and use more computational methods called machine learning which became prominent by the most recent AlphaGo and artificial intelligence issues. We present why this machine learning method is useful and necessary and how different theoretical linguistic explanation and computational linguistic feature representation are in terms of accessibilty to machine. We also show the method of representing linguistic features for the machine learning techniques i.e. neural network and bayesian network which started from human mind.

공기주입 장치와 공기제거 장치를 사용한 표면 플라즈몬 공명 타원계측기

이홍원,조은경,조재흥,원종명,신기량,제갈원,조용재,조현모,Lee, Hong-Won,Cho, Eun-Kyoung,Jo, Jae-Heung,Won, Jong-Myoung,Shin, Gi-Ryang,CheGal, Won,Cho, Yong-Jai,Cho, Hyun-Mo 한국광학회 2009 한국광학회지 Vol.20 No.3

완충용액과 분석시료의 확산으로 발생하는 측정 오차를 최소화하기 위하여 공기주입 장치와 공기제거 장치를 설치한 표면 플라즈 몬 공명 타원계측기(surface plasmon resonance ellipsometer: SPRE)를 새로이 제안하고 이를 제작하였다. SPRE에서 완충용액과 분석시료간의 상호 확산은 분석시료의 농도를 변화시켜 굴절률 측정에 영향을 주고, 이 결과 생체분자 물질의 접합 특성이 명확하게 나타나지 않는다. 이러한 SPRE의 측정 장치에 공기를 주입하는 장치를 추가로 설치하여 두 용액의 확산을 막고, 특히 불필요한 공기가 채널 내부로 유입되어 생기는 잡음신호를 없애기 위하여 공기제거 장치를 사용함으로써 신뢰성 있는 SPRE의 측정 결과를 얻을 수 있음을 확인하였다. The surface plasmon resonance ellipsometer (SPRE), using a multiple air injection system with an extraction of air system, has been proposed and developed to minimize measurement error of signals due to diffusion of reagent into running buffer. Since the diffusion of reagent into running buffer affects the refractive index of the running buffer by changing the concentration, characteristics of binding between various bio-molecules don't appear clearly in measurement results. The diffusion between running buffer and reagent can be blocked by using an air bubble injection system. An extraction of air system is used to remove the noise signal due to unnecessary air bubbles flowing in a channel. Reliability of measurement results has been improved by using the valve system.