• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.6
• /
• pp.235-242
• /
• 2021

In this work, we develop a Korean dependency parser based on a stack-pointer network that consists of a pointer network and an internal stack. The parser has an encoder and decoder and builds a dependency tree for an input sentence in a depth-first manner. The encoder of the parser encodes an input sentence, and the decoder selects a child for the word at the top of the stack at each step. Since the parser has the internal stack where a search path is stored, the parser can utilize information of previously derived subtrees when selecting a child node. Previous studies used only a grandparent and the most recently visited sibling without considering a subtree structure. In this paper, we introduce graph attention networks that can represent a previously derived subtree. Then we modify our parser based on the stack-pointer network to utilize subtree information produced by the graph attention networks. After training the dependency parser using Sejong and Everyone's corpus, we evaluate the parser's performance. Experimental results show that the proposed parser achieves better performance than the previous approaches at sentence-level accuracies when adopting 2-depth graph attention networks.

• KIISE Transactions on Computing Practices
• /
• v.24 no.3
• /
• pp.113-121
• /
• 2018

Position encoding is a method of applying weights according to position of words that appear in a sentence. Pointer networks is a deep learning model that outputs corresponding index with an input sequence. This model can be applied to coreference resolution using attribute. However, the pointer networks has a problem in that its performance is degraded when the length of input sequence is long. To solve this problem, we proposed two contributions to resolve the coreference. First, we applied position encoding and dynamic position encoding to pointer networks. Second, we stack deeply layers of encoder to make high-level abstraction. As results, the position encoding based stacked pointer networks model proposed in this paper had a CoNLL F1 performance of 71.78%, which was improved by 6.01% compared to vanilla pointer networks.

• Annual Conference on Human and Language Technology
• /
• 2018.10a
• /
• pp.13-18
• /
• 2018

구문 분석은 문장의 구조를 이해하며 의미의 중의성을 해결하는 것이다. 일반적으로 한국어는 어순 배열의 자유도가 높고 문장 성분의 생략이 빈번한 특성이 있기 때문에 의존 구문 분석이 주된 연구 대상이 되어 왔다. 스택-포인터 네트워크 모델은 의존 구문 파서에 맞게 포인터 네트워크 모델을 확장한 것이다. 스택-포인터 네트워크는 각 단어에서 의존소를 찾는 하향식 방식의 모델로 기존 모델의 장점을 유지하면서 각 단계에서 파생된 트리 정보도 사용한다. 본 연구에서는 스택-포인터 네트워크 모델을 한국어에 적용해보고 이와 함께 어절 정보를 반영하는 방법을 제안한다. 모델의 실험 결과는 세종 구문 구조를 중심어 후위(head-final)를 엄격히 준수하여 의존 구문 구조로 변환한 것을 기준으로 UAS 92.65%의 정확도를 얻었다.