CAJ | 학술논문

描述了一个基于依存关系的语义角色标注系统,该系统把依存关系作为语义角色标注的基本单元.通过手工或自动标注出来的依存关系,构造出依存关系树,并从树上抽取特征.用最大熵模型对句中谓词的语义角色进行识别和分类.为了消除不必要的结构化信息,在预处理阶段,依存关系树经过了Xue的剪枝算法处理.通过特征工程,丰富的特征及其组合被应用于系统.最终使用 CoNLL 2008 shared task提供的数据作为训练、开发和测试集,使用手工标注的依存关系,F1值达到了86.25%;使用MSTParser自动产生的依存关系,F1值达到了81.66%.
묘술료일개기우의존관계적어의각색표주계통,해계통파의존관계작위어의각색표주적기본단원.통과수공혹자동표주출래적의존관계,구조출의존관계수,병종수상추취특정.용최대적모형대구중위사적어의각색진행식별화분류.위료소제불필요적결구화신식,재예처리계단,의존관계수경과료Xue적전지산법처리.통과특정공정,봉부적특정급기조합피응용우계통.최종사용 CoNLL 2008 shared task제공적수거작위훈련、개발화측시집,사용수공표주적의존관계,F1치체도료86.25%;사용MSTParser자동산생적의존관계,F1치체도료81.66%.
This paper presents a dependency relations-based semantic role labeling system,which takes dependency relations as the labeled units.Here,a dependency tree is created from a sequence of labeled dependency relations,from which a number of features are extracted.Moreover,the maximum entropy model is trained to identify and classify the predicates' semantic roles.In the pre-processing step,a dependency tree is pruned using Xue's algorithm.With extensive feature engineering,a wide range of features and their combinations are applied.Evaluation on the CoNLL 2008 shared task shows that this method achieves the F1-score of 86.25% on the manually labeled version and the F1-score of 81.66% on the automatically labeled version using MSTParser.