• 인지과학
• /
• 제24권4호
• /
• pp.301-321
• /
• 2013

본 논문은 텍스트의 응집도 측정과 텍스트 자동생성 시스템을 위한 기초기술 중 하나인 문장정렬 과제에 대한 연구로, 개체기반적(entity-based) 접근의 한 유형인 개체격자 모형(Entity-Grid model)과 벡터공간 모형에 기반한 LSA(Latent Semantic Analysis)를 모두 시도하고 결과를 서로 비교하였다. 개체격자 모형에 대한 기존 연구들에서 논의된 명사들의 통사역(syntactic role) 정보가 한국어 텍스트 정렬과제에 미치는 영향을 실험하고자 하였으며, 기존 독일어권 응용연구 결과와는 달리 긍정적인 결과를 얻었다. 이 과정에서 한국어의 격조사를 활용하는 전략을 취했으며, 이는 한국어의 격표지 정보가 한국어 텍스트의 응집성을 측정하는 데에 유용할 수 있다는 점을 보인 것이다. 그리고 개체격자 모형을 통한 결과를 LSA 기반 모형결과와 비교하여 양 모형의 장단점과 향후 개선점을 아울러 논의하였다.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2001년도 하계종합학술대회 논문집(3)
• /
• pp.203-206
• /
• 2001

In the EKMT system, the part of Korea generation makes Korea sentence by using information obtained in the part of transfer. In the case of Korea generation, the conventional EKMT system don't arrange hierarchical word order and performs word order in the only modifier word. This paper proposes Korean adverb odering rule in English-Korean Machine Translation system which generates Korean sentence.

• 한국언어정보학회지:언어와정보
• /
• 제10권2호
• /
• pp.47-65
• /
• 2006

In this paper I address the problems of defining the semantics of numerals and accounting for how pragmatic inferences are made. I basically assume that a numeral n simply means '${\lambda}P{\lambda}x[#(x)n\;&\;P(x)]$', as commonly assumed. Even when a numeral n has 'at least' interpretation, a sentence with the number does not entail a sentence with n replaced with n-1. But when a sentence with n-1 holds, it is possible that a sentence with n or a larger number holds too. This is not based on a semantic relation, but on pragmatic informativeness. In addition to pragmatic strength, the actual reading of a numeral is affected by some background knowledge of generalizations about the world, but the ordering of pragmatic strength among numbers always plays a role in determining unilateral interpretations. In such a case, we can assume that a set of numbers relevant in the context forms a scale. Forming a scale does not necessarily lead to a unilateral interpretation. The bilateral interpretation of a number is possible in the context where it is known whether or not alternative sentences with contextually salient alternative numbers are true.

• Journal of Multimedia Information System
• /
• 제5권3호
• /
• pp.209-214
• /
• 2018

Today is the era of artificial intelligence. With the development of artificial intelligence, machines have begun to impersonate various human characteristics today. Chatbot is one instance of this interactive artificial intelligence. Chatbot is a computer program that enables to conduct natural conversations with people. As mentioned above, Chatbot conducted conversations in text, but Chatbot, in this study evolves to perform commands based on speech-recognition. In order for Chatbot to perfectly emulate a human dialogue, it is necessary to analyze the sentence correctly and extract appropriate response. To accomplish this, the sentence is classified into three types: objects, actions, and preferences. This study shows how objects is analyzed and processed, and also demonstrates the possibility of evolving from an elementary model to an advanced intelligent system. By this study, it will be evaluated that speech-recognition based Chatbot have improved order-processing time efficiency compared to text based Chatbot. Once this study is done, speech-recognition based Chatbot have the potential to automate customer service and reduce human effort.

• 한국언어정보학회:학술대회논문집
• /
• 한국언어정보학회 2008년도 정기학술대회
• /
• pp.138-146
• /
• 2008

Frequency adverbs can be interpreted as an adverb of quantification, and also as a frequentative adverb. These interpretations are related to the frequency adverbs' distributions, and the relation between semantics and syntax of frequency adverbs can be observed more explicitly when they appear with some other expressions in a sentence. Two frequency adverbs in Korean, cacwu and cakkwu, which seem to mean 'often/frequently', will be dealt with. We will specify their syntactic position by their interpretations derived from the relative ordering with other elements.

• 음성과학
• /
• 제12권4호
• /
• pp.129-139
• /
• 2005

This paper investigates any possible correlation between the information structure and the occurrence of clause internal pauses in the spontaneous discourse. One of the possible functions of pause is its capacity to signal the information structure of the discourse. However, this aspect was not much explored in Korean spontaneous speech. In the present study, information structure of spontaneous speech was defined for each word or word group on the basis of the information structure analysis model proposed by Van Donzel (1999) and Roulet (1991, 1997). Thus, at a local level (words or word groups) of discourse structure, a distinction was made between three types of information, new, given and inferable. The results showed that clause internal pauses tend to appear more frequently before new information than other types of information. However compared to the total number of words or word groups it was not noticed any specific ordering concerning different kind of information status and pausing. It was however found that clause internal pauses did not appear randomly. The majority of them occurred at the initial part of the clause or the sentence. This tendency was mostly related to the division of sentence (or clause) into topic and comment. Thus, the role of pauses as a marker of information structure seems to be less effective in spontaneous discourse.

• 한국음향학회지
• /
• 제41권3호
• /
• pp.359-366
• /
• 2022

본 논문에서는 콘포머 기반 FastSpeech2를 이용한 한국어 메뉴 음성합성기를 제안한다. 콘포머는 본래 음성 인식 분야에서 제안된 것으로, 합성곱 신경망과 트랜스포머를 결합하여 광역과 지역 정보를 모두 잘 추출할 수 있도록 한 구조다. 이를 위해 순방향 신경망을 반으로 나누어 제일 처음과 마지막에 위치시켜 멀티 헤드 셀프 어텐션 모듈과 합성곱 신경망을 감싸는 마카론 구조를 구성했다. 본 연구에서는 한국어 음성인식에서 좋은 성능이 확인된 콘포머 구조를 한국어 음성합성에 도입하였다. 기존 음성합성 모델과의 비교를 위하여 트랜스포머 기반의 FastSpeech2와 콘포머 기반의 FastSpeech2를 학습하였다. 이때 데이터셋은 음소 분포를 고려한 자체 제작 데이터셋을 이용하였다. 특히 일반대화 뿐만 아니라, 음식 주문 문장 특화 코퍼스를 제작하고 이를 음성합성 훈련에 사용하였다. 이를 통해 외래어 발음에 대한 기존 음성합성 시스템의 문제점을 보완하였다. ParallelWave GAN을 이용하여 합성음을 생성하고 평가한 결과, 콘포머 기반의 FastSpeech2가 월등한 성능인 MOS 4.04을 달성했다. 본 연구를 통해 한국어 음성합성 모델에서, 동일한 구조를 트랜스포머에서 콘포머로 변경하였을 때 성능이 개선됨을 확인하였다.

• 한국정보전자통신기술학회논문지
• /
• 제12권4호
• /
• pp.440-448
• /
• 2019

이 논문에서는 빈칸 되묻기 방식 기반 다중 키워드 처리가 가능한 주문용 챗봇을 제안한다. 일반적으로 챗봇을 이용한 주문 서비스의 경우에는 개발자가 미리 정의한 순서에 따라서만 주문이 진행된다. 그리고 한번의 답변으로 들어올 수 있는 입력 정보가 정해져 있기 때문에 사용자에 따라 다른 입력을 고려하지 못한다. 이 연구에서는 이러한 문제를 해결하기 위해 빈칸 되묻기 방식을 사용하여 다중 키워드 동시 처리를 하고자 한다. 빈칸 되묻기 방식은 다음과 같이 진행된다. 첫번째, 각 주문 단계에서 입력 받아야 하는 정보를 저장할 수 있는 배열을 미리 만들어 둔다. 그리고 각 주문 단계별로 받을 수 있는 정보들을 키워드로 미리 지정한다. 두번째로, 입력된 문장에서 키워드를 추출하는 작업을 진행한다. 그리고 추출된 키워드들을 해당하는 주문 단계의 배열에 채워 넣는다. 마지막으로, 각 주문 단계의 배열을 체크하면서 비어있는 단계에 대한 질문만 진행하여 부족한 정보들을 전부 채운다. 배열이 모두 채워지면 주문이 완료된다. 제안하는 방식은 한 문장에 주문과 관련된 키워드가 여러 개이더라도 처리가 가능하다. 그리고 한 번에 여러 개의 키워드를 처리할 수 있기 때문에 주문 단계를 생략하여 주문 시간을 줄일 수 있다. 안드로이드 스마트폰을 이용해 챗봇을 구현하고 빈칸 되묻기 방식을 이용해 주문 단계의 동적 처리가 되는지 실험을 통해 확인한다.

• 공학교육연구
• /
• 제15권5호
• /
• pp.131-139
• /
• 2012

This paper emphasizes writing education for engineering students should be communication driven writing education based on KEC2005. Communication driven writing for engineering students is essentially same as Technical Writing(TW) developed on the basis of ABET. Considering the current writing capability of engineering students and social need for various types of writing, TW education should be divided into two courses: basic and advanced. This paper deals with contents of a basic TW course in Myongji University, as a model case of a basic TW course for engineering students. It underlines various methods of prewriting that should be stressed and practiced in the TW class, because the prewriting step in the writing process determines the overall direction and structure of an essay. In particular, this paper introduces Power Writing(PW) which uses the structure of a paragraph as a means for providing building-blocks for the essay, employing logic, and ordering information arrangement in a paragraph. This paper also deals with important guidelines about sentence structure and word selection and proposes various applications of TW such as resume, interview, proposal, report, and presentation as a latter part of the basic course. Finally this paper highlights the etics of writing, such as plagiarism and the basic principles of quotation.

• 한국영어학회지:영어학
• /
• 제4권2호
• /
• pp.117-149
• /
• 2004

The paper is to argue against the asymmetric CP-V2 analysis of Old English, according to which finite verbs invariably undergo movement into a clause-final T within subordinate clauses and reach the functional head C within main clauses. The asymmetric CP-V2 analysis, first of all, faces difficulty in explaining a wide range of post-verbal elements within subordinate clauses. To resolve the problem, the analysis has to abandon the obligatoriness of V-to-T movement or introduce various types of extraposition whose status is dubious as a legitimate syntactic operation. Obligatory V-to-T movement in Old English lacks conceptual justification as well. Crosslinguistic evidence reveals that morphological richness in verbal inflection cannot entail overt verb movement. Moreover, the operation is always string-vacuous under the asymmetric CP- V2 analysis and has no effect at the interfaces, in violation of the principle of economy. The distribution of Old English finite verbs in main clauses also undermines the asymmetric CP-V2 analysis. Conceptually speaking, a proper syntactic trigger cannot be confirmed to motivate obligatory verb movement to C. The operation not only gets little support from nominative Case marking, the distribution of expletives, or complementizer agreement but also requires the unconvincing stipulation that expletives as well as sentence-initial subjects result from string-vacuous topicalization. Finally, textual evidence testifies that Old English sometimes permits non-V2 ordering patterns, many of which remain unexplained under the asymmetric CP-V2 analysis.