• 한국정보과학회논문지:소프트웨어및응용
• /
• 제36권4호
• /
• pp.335-345
• /
• 2009

이 논문은 웹 문서의 이미지 캡션 추출을 위한 방법으로서 이미지와 캡션의 위치적 연관성과 본문과 캡션의 어휘적 유사성을 동시에 고려한 방법을 제안한다. 이미지와 캡션의 위치적 연관성은 거리와 방향 관점에서 캡션이 이미지에 상대적으로 어떻게 위치하고 있는지를 나타내며, 본문과 캡션의 어휘적 유사성은 이미지를 설명하고 있는 캡션이 어휘적으로 본문과 어느 정도 유사한지를 나타낸다. 이미지와 캡션을 독립적으로 고려한 자질만을 사용한 캡션 추출 방법을 기저 방법으로 놓고 제안하는 방법들을 추가적인 자질로 사용하여 캡션을 추출하였을 때, 캡션 추출 정확률과 캡션 추출 재현율이 모두 향상되며, 캡션 추출 F-measure가 약 28% 향상되었다.

• 한국정보과학회논문지:소프트웨어및응용
• /
• 제36권8호
• /
• pp.642-652
• /
• 2009

본 논문에서는 기존의 관계 추출 성능을 향상시키기 위해서 기존의 자질 기반 방법에서 추구하였던 개체 주변 문맥 다양성 정보의 추출 및 적용과 커널 기반 방법의 강점인 관계 인스턴스에 대한 구문 구조적 자질 정보의 통합 활용을 통한 확장된 혼합 커널을 제안한다. ACE RDC 코퍼스를 활용한 실험에서, 기존의 합성곱 구문 트리 커널 기반 혼합 커널을 기반으로 총 9 종류의 평면적 어휘 자질 집합을 정의하고 이를 적용함으로써 성능 향상에 기여하는 어휘 자질 유형을 파악할 수 있었으며, 적은 규모의 학습 집합으로도 현재 최고 수준의 성능에 필적하는 결과를 얻을 수 있었다. 결론적으로 관계 추출을 위한 세 가지 핵심 정보, 즉 개체 자질, 구문 구조적 자질, 주변 문맥 어휘 자질을 통합 적용하면 관계 추출의 성능을 향상시킬 수 있음을 알 수 있었다.

• 한국언어정보학회:학술대회논문집
• /
• 한국언어정보학회 2007년도 정기학술대회
• /
• pp.29-41
• /
• 2007

Concepts of greater and greater complexity can be constructed by building systems of entities, by relating other entities to that system with a figure-ground relation, by embedding concepts of figure-ground in the concept of change, by embedding that in causality, and by coarsening the granularity and beginning the process over again. This process can be called the Ontological Ascent. It pervades natural language discourse, and suggests that to do lexical semantics properly, we must carefully axiomatize abstract theories of systems of entities, the figure-ground relation, change, causality, and granularity. In this paper, I outline what these theories should look like.

• 대한음성학회지:말소리
• /
• 제60호
• /
• pp.29-45
• /
• 2006

We investigated whether the first syllable plays an important role in lexical access in Korean visual word recognition. To do so, one lexical decision task (LDT) and two form primed LDT experiments examined the nature of the syllabic neighborhood effect. In Experiment 1, the syllabic neighborhood density and the syllabic neighborhood frequency was manipulated. The results showed that lexical decision latencies were only influenced by the syllabic neighborhood frequency. The purpose of experiment 2 was to confirm the results of experiment 1 with form-primed LDT task. The lexical decision latency was slower in form-related condition compared to form-unrelated condition. The effect of syllabic neighborhood density was significant only in form-related condition. This means that the first syllable plays an important role in the sub-lexical process. In Experiment 3, we conducted another form-primed LDT task manipulating the number of syllabic neighbors in words with higher frequency neighborhood. The interaction of syllabic neighborhood density and form relation was significant. This result confirmed that the words with higher frequency neighborhood are more inhibited by neighbors sharing the first syllable than words with no higher frequency neighborhood in the lexical level. These findings suggest that the first syllable is the unit of neighborhood and the unit of representation in sub-lexical representation is syllable in Korea.

• 한국정보과학회논문지:소프트웨어및응용
• /
• 제35권10호
• /
• pp.627-635
• /
• 2008

본 연구에서는 사용자 어휘지능망(U-WIN)의 어휘 관계 중의 하나인 구문관계를 자동으로 구축하는 방법을 제시하고자 한다. 먼저, 구문관계를 형성할 수 있는 후보명사를 용언의 용례에서 문형 정보를 기준으로 추출함으로써, 용언의 세분화된 의미별로 정확하고 다양한 후보명사를 추출할 수 있다. 그러나 추출된 후보명사는 다양한 의미를 지니고 있으므로, 어휘간의 명확한 구문관계를 설정하기 위해서는 후보명사의 여러 의미 중에서 정확한 의미로 결정해야 한다. 그래서 본 연구에서는 용례 매칭 규칙, 구문 패턴, 의미 유사도, 빈도 정보 등을 이용하여 후보명사의 의미를 분별한다. 또한 구문패턴의 빈도 정보를 이용하여 용례에 나타나지 않지만 구문관계를 형성할 수 있는 명사를 추출하여 구문관계를 확장하고자 하였다. 이러한 연구는 명사 중심의 어휘망이 용언과의 구문관계 구축을 통해 형태소 분석, 구문 분석, 의미분석 등에 광범위하게 활용할 수 있는 어휘망의 기반을 다지는 작업이 될 수 있을 것이다.

• 인지과학
• /
• 제3권1호
• /
• pp.1-24
• /
• 1991

본 논문은 한국어의 어휘적 중의성을 해결하는 시스템을 보여준다. 이 시스템은 의존 문법과 대조 의미론을 이용하고 있다. 의존 문법은 두 형태소 사이의 의존관계에 의하여 문장을 분석한다. 대조 의미론은 어휘적 중의성과 의미관계의 상호작용을 조사한다. 대조 의미론은 의미 프레임,의미 백터,대조,분류의 4개의 구성요소로 이루어진다. 본 시스템은 C 언어로 구성되었으며, 문자을 분석 학과 두 단어간의 의미 관계를 조사하며 어휘적 중의성을 해결한다.

• 말소리와 음성과학
• /
• 제4권1호
• /
• pp.55-65
• /
• 2012

In this study, the prosodic disambiguation of the syntactic attachment differences was investigated in relation to the effect of lexical bias. Speech materials were composed of N1-conj-N2-PP phrases such as "walkers and runners with dogs." The results show that the use of durational pattern is dominant over the pitch pattern to differentiate the attachment differences. The characteristic pitch contour was the rise and fall over N1 and N2 in the high attachment. The pitch contour in the low attachment was the rise and fall over N2 and N3 although the frequency of such patterns was lower for the low attachment case. For the durational pattern, the lengthening in the N2 region plays a significant role in the disambiguation of the syntactic attachments. The interaction between the lexical bias and the syntactic attachment was not statistically significant in the duration data.

• 말소리와 음성과학
• /
• 제7권4호
• /
• pp.49-55
• /
• 2015

It is argued in this paper that, in determining semantic similarity, Korean words should be recategorized with a focus on the semantic relation to ontology in light of cross-linguistic morphological variations. It is proposed, in particular, that Korean semantic similarity should be measured on three tracks, human judgements track, relatedness track, and cross-part-of-speech relations track. As demonstrated in Yang et al. (2015), GloVe, the unsupervised learning machine on semantic similarity, is applicable to Korean with its performance being compared with human judgement results. Based on this compatability, it was further thought that the model's performance might most likely vary with different kinds of specific relations in different languages. An attempt was made to analyze them in terms of two major Korean-specific categories involved in their lexical and cross-POS-relations. It is concluded that languages must be analyzed by varying methods so that semantic components across languages may allow varying semantic distance in the vector space models.

• 한국주거학회논문집
• /
• 제17권4호
• /
• pp.131-143
• /
• 2006

Nowadays, diverse mass media discovers and introduces outstanding management cases of apartment complexes to induce vital competitions of constructors and active participation of residents to apartment management. This study statistically analyzed the management issues of outstanding apartment complexes that have been introduced by mass media with lexical criteria to examine the characteristics of their exemplary management. The key issues of outstanding apartment management are summarized as: efficient management of convenient facilities for residents, community activities based on residents' participation, and maintenance of pleasant living environments through transparent management. Also, the result of the relation arrangement of co-occurrence word from a Social Network Analysis included three key concepts of multi-family housing management - Maintenance Management, Operating Management, and Community Life Management - with emphasis on 'residents' and 'apartment complexes.' However, Operating Management was relatively deemphasized.

• 한국영어학회지:영어학
• /
• 제1권1호
• /
• pp.101-117
• /
• 2001

In English, whether or not wh-movement creates weak crossover effects depends upon the type of wh-phrases that cross over. A bare interrogative like who shows a typical weak crossover effect whereas which N type (e.g. which girl) and partitive type (e.g. which of these girls) wh-phrases would show mere weaker and weakest crossover effects, respectively. Previous approaches to English crossover phenomena that resort to a binary notion of specificity or D-linking cannot account for the three-way contrast the three different types of wh-phrases show. To overcome this problem, I argue in this paper that specificity should be a non-binary set-theoretic notion and propose the following subset principle and optimal binding relation: Between two lexical nominal expressions A and B, A is regarded as more specific than B iff the denotation of A comes from a more narrowly defined non-singleton set than B. Between two lexical nominal expressions A and B, if A locally binds B, then the non-singleton set from which the denotation of A comes should be a subset of the set from which the denotation of B comes (i.e. B cannot be more specific than A). The smaller the subset (i.e. the wider the specificity gap between binder and binder), the more optimal the local binding relation is. A locally binds B iff A is coindexed with B, and A c-commands B, and there is no such C that does not bind A but binds B. Finally, I show that partitivity functions to carve out a smaller subset and thus make partitive wh-phrases more specific than simple which N type wh-phrases.