CAJ | 학술논문

事件抽取是自动内容抽取(Automatic Content Extraction,ACE)会议评测的任务之一,事件要素识别是事件抽取的一个子任务.分析了事件抽取和事件要素识别的研究现状,提出了一种基于特征加权的事件要素识别算法(Feature Weighting Based Event Argument Identification,FWEAI).该算法首先对分类算法中的ReliefF特征选择算法进行改进,将其应用于聚类算法中.改进的ReliefF算法(FWA)根据各个特征对聚类的不同贡献分配不同的权值,然后采用KMeans算法对事件要素进行聚类.实验结果表明,FWEAI算法可以提高事件要素识别的准确率.
사건추취시자동내용추취(Automatic Content Extraction,ACE)회의평측적임무지일,사건요소식별시사건추취적일개자임무.분석료사건추취화사건요소식별적연구현상,제출료일충기우특정가권적사건요소식별산법(Feature Weighting Based Event Argument Identification,FWEAI).해산법수선대분류산법중적ReliefF특정선택산법진행개진,장기응용우취류산법중.개진적ReliefF산법(FWA)근거각개특정대취류적불동공헌분배불동적권치,연후채용KMeans산법대사건요소진행취류.실험결과표명,FWEAI산법가이제고사건요소식별적준학솔.
Vent extraction is a task of Automatic Content Extraction (ACE) program. Event Argument Identification is sub-task of Event extraction. The state-of-the-art of event extraction and event argument identification was given. An algorithm named Feature Weighting Based Event Argument Identification (FWEAI) was proposed, which improved ReliefF, a feature selection algorithm in classification algorithm, and employed it in clustering algorithm. The improved ReliefF (FWA) can assign different weights to different features according to their contributions to clustering, then Event arguments were clustered by using KMeans clustering algorithm. Experimental results demonstrate that FWEAI algorithm improves the precision on Event Argument Identification.