• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.235-240
• /
• 2021

구문 트리의 구조적 정보는 문장 수준 관계 추출을 수행하는데 있어 매우 중요한 자질 중 하나다. 기존 관계 추출 연구는 구문 트리에서 최단 의존 경로를 적용하는 방식으로 관계 추출에 필요한 정보를 추출해서 활용했다. 그러나 이런 트리 가지치기 기반의 정보 추출은 관계 추출에 필요한 어휘 정보를 소실할 수도 있다는 문제점이 존재한다. 본 논문은 이 문제점을 해소하기 위해 개체 중심으로 구문 트리를 재구축하고 모든 노드의 정보를 관계 추출에 활용하는 모델을 제안한다. 제안 모델은 TACRED에서 F1 점수 74.9 %, KLUE-RE 데이터셋에서 72.0%로 가장 높은 성능을 보였다.

• Journal of KIISE:Software and Applications
• /
• v.28 no.11
• /
• pp.822-832
• /
• 2001

We have implemented an inductive learning system that learns PROSPECTOR-rule-style classification rules from sets of examples. In our a approach, a genetic algorithm is used in which a population consists of rule-sets and rule-sets generate offspring through the exchange of rules relying on genetic operators such as crossover, mutation, and inversion operators. In this paper, we describe our learning environment centering on the syntactic structure and meaning of classification rules, the structure of a population, and the implementation of genetic operators. We also present a method to evaluate the performance of rules and a heuristic approach to generate rules, which are developed to implement mutation operators more efficiently. Moreover, a method to construct a classification system using multiple learned rule-sets to enhance the performance of a classification system is also explained. The performance of our learning system is compared with other learning algorithms, such as neural networks and decision tree algorithms, using various data sets.

• Journal of KIISE
• /
• v.41 no.11
• /
• pp.992-1005
• /
• 2014

Coreference resolution finds all expressions that refer to the same entity in a document. Coreference resolution is important for information extraction, document classification, document summary, and question answering system. In this paper, we adapt Stanford's Multi-pass sieve system, the one of the best model of rule based coreference resolution to Korean. In this paper, all noun phrases are considered to mentions. Also, unlike Stanford's Multi-pass sieve system, the dependency parse tree is used for mention extraction, a Korean acronym list is built 'dynamically'. In addition, we propose a method that calculates weights by applying transitive properties of centers of the centering theory when refer Korean pronoun. The experiments show that our system obtains MUC 59.0%, $B_3$ 59.5%, Ceafe 63.5%, and CoNLL(Mean) 60.7%.