摘要
事件事实性指出了事件发生与否的确定性程度,是自然语言理解的基础。在研究过程中,针对中文事件的事实性识别问题,提出了一种基于特征工程的有效识别方法。该方法选取事件的事实性相关信息进行特征的处理和转化。同时,考虑到部分特征与事件事实性之间的联系,依据规则进行特征融合。实验证明,相比基于规则的事件事实性识别方法,该方法有着更好的识别效果。
Event factuality refers to the level of event factual information expressed by event narrator and it is the foun- dation of natural language understanding. During the research, we focused on identifying Chinese event factuality and proposed an effective identification approach based on features. It extracts and transforms features from the factual re- lated information. Meanwhile, considering the relationship between parts of features and event factuality, it makes a fu- sion of these features according to rules. Experimental results manifest that our approach achieves a higher performance than the rule-based approach for the task of event factuality identification.
出处
《计算机科学》
CSCD
北大核心
2017年第5期241-244,256,共5页
Computer Science
基金
国家自然科学基金(61472265)
国家自然科学基金重点项目(61331011)
江苏省前瞻性联合研究项目(BY2014059-08)
软件新技术与产业化协同创新中心部分资助
关键词
事件事实性
识别方法
特征的转化和融合
Event factuality, Identification approach, Feature transformation and fusion
