• Annual Conference on Human and Language Technology
• /
• 2009.10a
• /
• pp.6-13
• /
• 2009

본 연구에서는 인간언어공학에서의 활용을 위해 세종전자사전의 의미부류와 KorLexNoun 1.5의 상위노드 간의 사상을 목표로 전문가의 수작업에 의한 세밀한 사상 방법론(fine-grained mapping method)을 제안한다. 또한 이질적인 두 이종 자원 간의 사상에 있어 각 의미체계의 이질성으로 인해 발생하는 여러 가지 문제점을 살펴보고, 그 해결방안을 제안한다. 본 연구는 세종의미부류체계가 밝히고자 했던 한국어의 의미구조와, Prinston WordNet을 참조로 하여 KorLexNoun에 여전히 영향을 미치고 있는 영어 의미구조를 비교함으로써 공통점과 차이점을 파악할 수 있고, 이를 바탕으로 언어 독립적인 개념체계를 구축하는 데 기여할 수 있다. 또한 향후 KorLex의 용언에 기술되어 있는 문형정보와 세종 전자사전의 용언의 격틀 정보를 통합 구축하여 구문분석에서 이용할 때, 세종 의미부류와 KorLexNoun의 상위노드를 통합 구축함으로써 논항의 일반화된 선택제약규칙의 기술에서 이용될 수 있다. 본 연구에서 제안된 사상방법론은 향후 이종 자원의 자동 사상 연구에서도 크게 기여할 것이다. 아울러 두 이종 자원의 사상을 통해 두 의미체계가 지닌 장점을 극대화하고, 동시에 단점을 상호 보완하여 보다 완전한 언어자원으로써 구문분석이나 의미분석에서 이용될 수 있다.

• Proceedings of the Korea Contents Association Conference
• /
• 2006.11a
• /
• pp.570-574
• /
• 2006

KISTI portal service is currently presenting the documents with many terminologies, so users can't find the results having their intention by using an umbrella query. In this paper, we suggest user oriented retrieval service that reflects query auto-complete, related-word suggestion and query expansion that uses nouns and relationships of U-WIN which is known as a semantic language resource. We intend to advance the retrieval satisfaction of current science & technology information service by using U-WIN's semantic information and improve the service environment that user can retrieve what they want quickly and exactly.

• Proceedings of the Korean Society for Language and Information Conference
• /
• 2005.06a
• /
• pp.71-85
• /
• 2005

To build a Korean wordnet, we translated semi-automatically the English wordnet PWN into Korean verbs. During this process, we found that some of translation errors are related to the arbitrariness of PWN`s sense distinction in regard to the accusativity/inaccusativity of the same verb form in English. This study presents an empirically based method of remodeling the PWN for Korean wordnet and while revising the PWN`s hierarchical structure, it shows the necessity of classifying the Korean passive verbs as semantically autonomous.

• 재활복지
• /
• v.21 no.3
• /
• pp.187-207
• /
• 2017

The purpose of this study was to explore the characteristics of defining concrete and abstract nouns for the elderly. A total of 382 elderly participated in this study and they were classified into four age groups (i.e., over 55 to under 64, over 65 to under 74, over 75 to under 84, and over 85 year-old group). They performed the word definition task, composed of five concrete and five abstract nouns. The total scores and numbers and ratio of core/supplementary meanings were compared among four elderly groups. The frequency and ratio of error types were also examined. The results showed that all four groups had statistically significant differences in total scores, numbers and ratio of core and supplementary meaning of concrete noun definition task. In addition, abstract noun definition performances revealed group differences except the two groups (over 75 to under 84 and over 85-year-old group). The oldest group showed a sharp increase in error production. The highest ratio of error types were personal experience in over 55 to under 64-year-old group, and over 65 to under 74 year-old groups; and for the target word repetition in over 75 to under 84 year-old group; and no response in over 85 year-old group. In conclusion, both concrete and abstract word defining abilities had age-related deterioration. This decline results from impairment in spreading semantic knowledge within semantic network, which is vulnerable to aging. Characteristics of word definition for elderly can provide basic information to understand various neurolinguistic disorders associated with age.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.42 no.3
• /
• pp.428-437
• /
• 2018

This study derives evaluation terms by analyzing the semantic relationship between design elements and sentiment terms in regards to fashion design. As for research methods, a total of 38,225 texts from Daum and Naver Blogs from November 2015 to October 2016 were collected to analyze the parts, frequency, centrality and semantic networks of the terms. As a result, design elements were derived in the form of a noun while fashion image and user's emotional responses were derived in the form of adjectives. The study selected 15 noun terms and 52 adjective terms as evaluation terms for men's striped shirts. The results of semantic network analysis also showed that the main contents of the users of men's striped shirts were derived as characteristics of expression, daily wear, formation, and function. In addition, design elements such as pattern, color, coordination, style, and fit were classified with evaluation results such as wide, bright, trendy, casual, and slim.

• Journal of Intelligence and Information Systems
• /
• v.12 no.1
• /
• pp.107-123
• /
• 2006

We propose the Generalized Binary Second-order Recurrent Neural Networks(GBSRNNf) being equivalent to regular grammars and ?how the implementation of lexical analyzer recognizing the regular languages by using it. All the equivalent representations of regular grammars can be implemented in circuits by using GSBRNN, since it has binary-valued components and shows the structural relationship of a regular grammar. For a regular grammar with the number of symbols m, the number of terminals p, the number of nonterminals q, and the length of input string k, the size of the corresponding GBSRNN is $O(m(p+q)^2)$ and its parallel processing time is O(k) and its sequential processing time, $O(k(p+q)^2)$.

• Language and Information
• /
• v.14 no.1
• /
• pp.165-196
• /
• 2010

The primary goal of this paper is to propose methods of enriching two heterogeneous ontologies: Sejong Semantic Classes (SJSC) and KorLexNoun 1.5 (KLN). In order to achieve this goal, this study introduces the pros and cons of two ontologies, and analyzes the error patterns found during the fine-grained manual mapping processes between them. Error patterns can be classified into four types: (1) structural defectives involved in node branching, (2) errors in assigning the semantic classes, (3) deficiency in providing linguistic information, and (4) lack of the lexical units representing specific concepts. According to these error patterns, we propose different solutions in order to correct the node branching defectives and the semantic class assignment, to complement the deficiency of linguistic information, and to increase the number of lexical units suitably allotted to their corresponding concepts. Using the results of this study, we can obtain more enriched ontologies by correcting the defects and errors in each ontology, which will lead to the enhancement of practicality for syntactic and semantic analysis.

• The KIPS Transactions:PartB
• /
• v.18B no.6
• /
• pp.405-414
• /
• 2011

Korean grammar checkers typically detect context-dependent errors by employing heuristic rules that are manually formulated by a language expert. These rules are appended each time a new error pattern is detected. However, such grammar checkers are not consistent. In order to resolve this shortcoming, we propose new method for generalizing error decision rules to detect the above errors. For this purpose, we use an existing thesaurus KorLex, which is the Korean version of Princeton WordNet. KorLex has hierarchical word senses for nouns, but does not contain any information about the relationships between cases in a sentence. Through the Tree Cut Model and the MDL(minimum description length) model based on information theory, we extract noun classes from KorLex and generalize error decision rules from these noun classes. In order to verify the accuracy of the new method in an experiment, we extracted nouns used as an object of the four predicates usually confused from a large corpus, and subsequently extracted noun classes from these nouns. We found that the number of error decision rules generalized from these noun classes has decreased to about 64.8%. In conclusion, the precision of our grammar checker exceeds that of conventional ones by 6.2%.

• KIISE Transactions on Computing Practices
• /
• v.23 no.4
• /
• pp.226-231
• /
• 2017

Word representation, an important area in natural language processing(NLP) used machine learning, is a method that represents a word not by text but by distinguishable symbol. Existing word embedding employed a large number of corpora to ensure that words are positioned nearby within text. However corpus-based word embedding needs several corpora because of the frequency of word occurrence and increased number of words. In this paper word embedding is done using dictionary definitions and semantic relationship information(hypernyms and antonyms). Words are trained using the feature mirror model(FMM), a modified Skip-Gram(Word2Vec). Sense similar words have similar vector. Furthermore, it was possible to distinguish vectors of antonym words.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2013.06a
• /
• pp.401-404
• /
• 2013

청각장애인을 위한 지상파방송 서비스 중 자막방송은 100%에 가까운 편성 비율을 달성하고 있지만, 화면을 가리는 수화방송은 5% 수준의 편성에 그치고 있다. 본 연구에서는 자막방송을 수화로 번역하여 그래픽 수화방송을 생성함으로써 수화방송의 비율을 높이고자 한다. 수화 단어들의 빈도를 파악하고 중요 단어부터 모션 캡처하기 위해 과거 3년간 일기예보 스크립트를 분석하였다. 자막방송 문장을 형태소별로 분석한 다음 중요 품사 위주로 단어 단위로 번역하고, 기 구축된 한국어 어휘의미망을 이용하여 수화사전에 없는 유의어와 하위어를 대표어로 대체하였다. 기계번역 기술이 수화통역사의 수준을 따라잡을 수는 없지만 향후 수화방송도 선택적 서비스가 가능해지고 수화통역사의 수화방송이 모든 프로그램에 편성될 때까지 본 시스템이 보조적 시청 수단으로 사용 가능할 것이다.