• 한국콘텐츠학회논문지
• /
• 제12권1호
• /
• pp.17-29
• /
• 2012

급격하게 성장하고 있는 오픈 리소스인 링크드 데이터는 최근 선진국 정부의 많은 관심 속에 데이터 공개 및 상호운용성 확보를 위한 방안으로 주목받고 있다. 그러나 신뢰할 수 있는 개체 식별 기술의 부재로 링크드 데이터의 양적 성장에 비해 개체 수 대비 링크의 수가 적은 현상과 일부 데이터 셋에 링크가 집중되는 현상을 보이고 있다. 본 연구에서는 이러한 링크드 데이터의 문제를 해결하기 위해 개체 간 관계(owl:sameAs, owl differentFrom 등)를 이용하거나 Curation 방식을 사용하는 기존 링크드 데이터 기반 개체 식별 방식의 문제를 다중 온톨로지의 개체 식별이 가능한 자동화된 개체 식별 방식을 통해 개선하고 저자 개체의 대응 속성과 개체 유형의 논리적 특성을 활용하여 개체 식별 정합성을 검증할 수 있는 다중 온톨로지 기반의 실시간 저자 식별 방법을 제안하고 평가한다. 본인의 확인을 거친 29명의 저자 정보를 이용해 개체 식별 정확성 결과를 평가하여 평균 0.8533 (K measure)의 긍정적인 성능을 보였다.

• ETRI Journal
• /
• 제43권6호
• /
• pp.1038-1048
• /
• 2021

We propose an end-to-end neural coreference resolution for the Korean language that uses an attention mechanism to point to the same entity. Because Korean is a head-final language, we focused on a method that uses a pointer network based on the head. The key idea is to consider all nouns in the document as candidates based on the head-final characteristics of the Korean language and learn distributions over the referenced entity positions for each noun. Given the recent success of applications using bidirectional encoder representation from transformer (BERT) in natural language-processing tasks, we employed BERT in the proposed model to create word representations based on contextual information. The experimental results indicated that the proposed model achieved state-of-the-art performance in Korean language coreference resolution.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• 제12권4호
• /
• pp.277-284
• /
• 2012

Named entity recognition (NER) technique can play a crucial role in extracting information from the web. While NER systems with relatively high performances have been developed based on careful manipulation of terms with a statistical model, term mismatches often degrade the performance of such systems because the strings of all the candidate entities are not known a priori. Despite the importance of lexical-level term mismatches for NER systems, however, most NER approaches developed to date utilize only the term string itself and simple term-level features, and do not exploit the semantic features of terms which can handle the variations of terms effectively. As a solution to this problem, here we propose to match the semantic concepts of term units in restaurant named entities (NEs), where these units are automatically generated from multiple resolutions of a semantic tree. As a test experiment, we applied our restaurant NER scheme to 49,153 nouns in Korean restaurant web pages. Our scheme achieved an average accuracy of 87.89% when applied to test data, which was considerably better than the 78.70% accuracy obtained using the baseline system.

• 정보처리학회논문지D
• /
• 제11D권3호
• /
• pp.553-562
• /
• 2004

본 논문에서는 데이터베이스 설계시 중요한 과정인 개념설계 단계에서 개체관계도(Entity Relationship Diagram)의 정보를 저장하는 ERD시소러스(Thesaurus)를 구축하고, 이러한 ERD시소러스를 기반으로 하는 뷰 통합 방법론을 제시하고자 한다. 제시된 방법론의 유용성을 입증하기 위하여, 적용사례에 대한 뷰 통합지원시스템의 프로토타입을 구축하였다. 적용결과, ERD시소러스 기반의 방법론이 기존의 뷰 통합 방법론보다 친밀도 분석, 의미충돌 해결, 유 통합과정에서 더 효과적임을 확인할 수 있었다. 따라서 이 방법론이 기존의 단편화된 스키마의 통합이나 대규모 데이터베이스 통합 설계시 유용하게 활용될 수 있을 것으로 기대된다.

• 정보과학회 논문지
• /
• 제44권5호
• /
• pp.496-502
• /
• 2017

포인터 네트워크(Pointer Networks)는 Recurrent Neural Network (RNN)를 기반으로 어텐션 메커니즘(Attention mechanism)을 이용하여 입력 시퀀스에 대응되는 위치들의 리스트를 출력하는 딥 러닝 모델이다. 대명사 상호참조해결은 문서 내에 등장하는 대명사와 이에 대응되는 선행사를 찾아 하나의 엔티티로 정의하는 자연어처리 문제이다. 본 논문에서는 포인터 네트워크를 이용하여 대명사와 선행사의 참조관계를 밝히는 대명사 상호참조해결 방법과 포인터 네트워크의 입력 연결순서(chaining order) 여섯가지를 제안한다. 실험 결과, 본 논문에서 제안한 방법 중 연결순서 coref2 가 MUC F1 81.40%로 가장 좋은 성능을 보였다. 이는 기존 한국어 대명사 상호참조해결의 규칙 기반(50.40%)보다 31.00%p, 통계 기반(62.12%) 보다 19.28%p 우수한 성능임을 나타낸다.

• 정보과학회 논문지
• /
• 제41권11호
• /
• pp.992-1005
• /
• 2014

상호참조해결은 문서 내에서 선행하는 명사구와 현재 등장한 명사구 간에 같은 개체를 의미하는 지를 결정하는 문제로 정보 추출, 문서분류 및 요약, 질의응답 등에 적용된다. 본 논문은 상호참조해결의 규칙기반 방법 중 가장 성능이 좋은 Stanford의 다 단계 시브(Multi-pass Sieve) 시스템을 한국어에 적용한다. 본 논문에서는 모든 명사구를 멘션(mention)으로 다루고 있으며, Stanford의 다 단계 시브 시스템과는 달리 멘션 추출을 위해 의존 구문 트리를 이용하고, 동적으로 한국어 약어 리스트를 구축한다. 또한 한국어 대명사를 참조하는데 있어 중심화 이론 중 중심의 전이적인 특성을 적용하여 가중치를 부여하는 방법을 제안한다. 실험 결과 F1 값은 MUC 59.0%, B3 59.5%, Ceafe 63.5%, CoNLL(평균) 60.7%의 성능을 보였다.

• ETRI Journal
• /
• 제45권1호
• /
• pp.93-104
• /
• 2023

Coreference resolution is a task in discourse analysis that links several headwords used in any document object. We suggest pointer networks-based coreference resolution for Korean using multi-task learning (MTL) with an attention mechanism for a hierarchical structure. As Korean is a head-final language, the head can easily be found. Our model learns the distribution by referring to the same entity position and utilizes a pointer network to conduct coreference resolution depending on the input headword. As the input is a document, the input sequence is very long. Thus, the core idea is to learn the word- and sentence-level distributions in parallel with MTL, while using a shared representation to address the long sequence problem. The suggested technique is used to generate word representations for Korean based on contextual information using pre-trained language models for Korean. In the same experimental conditions, our model performed roughly 1.8% better on CoNLL F1 than previous research without hierarchical structure.

• 대순사상논총
• /
• 제21권
• /
• pp.99-156
• /
• 2013

This paper is to indicate research tendencies of faith in Daesoonjinrihoe and controversial points of those, and to consider the outlook on Sangje after defining it as theological understanding and explanation for Gu-Cheon-Sang-Je (High-est ruling Entity that is the object of devotion in Daesoon-jinrihoe). As the first introduction to the work, various descriptions for Sangje are arranged and the meanings of those are analyzed. In brief, first, the name of Gu-Cheon-Eung-Won-Nweh-Seong-Bo-Hwa-Cheon-Jon, expresses the fact that the authority of Sangje (the Supreme Entity) is exposed by spatial concept Sangje dwells in Ninth Heaven. This fact can be compared with the doctrines Allah in Islam and Jehovah in Christianity each are dwelled in Seventh Heaven. And the name shows Sangje is the ruler who reigns over the universe by using yin and yang. Second, the name, Gu-Cheon-Eung-Won-Nweh-Seong-BoHwa-Cheon-Jon, is imported from China Taoism because it has been in Ok-Chu-Gyeong (the Gaoshang shenlei yushu). But in fact it's root is in Korea because Buyeo and Goguryeo, the ancient Korean nations, have the source of the name. While the name is not the Supreme Entity in China Taoism, it is the Supreme Entity in Daesoonjinrihoe. This fact is a important difference. Third, arbitrarily or not, the name, Gu-Cheon-Eung-Won-Nweh-Seong-Bo-Hwa-Cheon-Jon, is put on the image of 'resolution of grievances'. The reason is that many peoples in Korea and China has called the name for about 1,000 years ago to help their fortunes and escape predicaments. Forth, not only Gu-Cheon-Eung-Won-Nweh-Seong-Bo-Hwa-Cheon-Jon but also the name, Three Pure Ones and Ok-Cheon-Jin-Wang (Yuqingzhenwang) in China Taoism used as the Highest ruling Entity in Daesoonjinrihoe. But the relations between three Pure Ones and Ok-Cheon-Jin-Wang and Gu-Cheon-Eung-Won-Nweh-Seong-Bo-Hwa-Cheon-Jon in Dae-soonjinrihoe are different from that in China Taoism. Fifth, Sangje is associated with the Polaris divinity of Tae-Eul, view on God in Oriental Cosmology. The description Tae-Eul as well as Gu-Cheon-Eung-Won-Nweh-Seong-Bo-Hwa-Cheon-Jon is indicated Sangje is linked to the faith of Buyeo and Goguryeo. Sixth, Sangje is not only Mugeuk-Sin (The God of The Endless) who supervise the Endless but also Taegeuk-Ji-Cheon-Jon (The God of The Ultimate Reality) who supervise the Ultimate Reality. These descriptions directly display the fact Sangje is a creator. Seventh, in case explaining Sangje, the point of view is necessary that grasps the whole viewpoints Sangje 'was' Hidden God(deus otiosus) and 'is' Unhidden God after Incarnation. Eighth, Sangje is Cheon-Ju in Donghak, but different from that. Cheon-Ju in Donghak has both transcendence and immanence in tightrope tension, but Cheon-Ju in Daesoonjinrihoe emphasize transcendence than immanence. That difference is the result of the fact Cheon-Ju in Donghak was a being having revealed a man and Cheon-Ju in Daesoonjinrihoe was a being having incarnated after revealing a man. Ninth, Sangje is Gae-Byeok-Jang who is the manager of the transforming and ordering the Three Realms of the World by the Great Do which is the mutual beneficence of all life and Hae-Won-Sin who is the God of resolution of grievances.

• Journal of Korean Neurosurgical Society
• /
• 제45권4호
• /
• pp.253-255
• /
• 2009

Spontaneous spinal subarachnoid hematoma (SSH) is a rare entity to cause spinal cord or nerve root compression and is usually managed as surgical emergencies. We report a case of spontaneous SSH manifesting as severe lumbago, which demonstrated nearly complete clinical resolution with conservative treatment A 58-year-old female patient developed a large SSH, which was not related to blood dyscrasia, anticoagulation, lumbar puncture. or trauma. Patient had severe lumbago but no neurologic deficits. Because of absence of neurologic deficits, she was treated conservatively. Follow-up magnetic resonance (MR) image showed complete resolution. Conservative treatment of SSH may be considered if the patient with spontaneous SSH has no neurologic deficits.

• ETRI Journal
• /
• 제46권1호
• /
• pp.35-47
• /
• 2024

This study introduces CR-M-SpanBERT, a coreference resolution (CR) model that utilizes multiple embedding-based span bidirectional encoder representations from transformers, for antecedent recognition in natural language (NL) text. Information extraction studies aimed to extract knowledge from NL text autonomously and cost-effectively. However, the extracted information may not represent knowledge accurately owing to the presence of ambiguous entities. Therefore, we propose a CR model that identifies mentions referring to the same entity in NL text. In the case of CR, it is necessary to understand both the syntax and semantics of the NL text simultaneously. Therefore, multiple embeddings are generated for CR, which can include syntactic and semantic information for each word. We evaluate the effectiveness of CR-M-SpanBERT by comparing it to a model that uses SpanBERT as the language model in CR studies. The results demonstrate that our proposed deep neural network model achieves high-recognition accuracy for extracting antecedents from NL text. Additionally, it requires fewer epochs to achieve an average F1 accuracy greater than 75% compared with the conventional SpanBERT approach.