民航领域突发事件的实体链接方法

实体链接的相关研究主要集中于医疗、生物和新闻领域，但在民航领域的研究较少。因此针对民航领域实体链接任务进行了研究，发现在实体链接中存在实体变体多、歧义少等问题。为解决上述问题，提出了一种基于语义推断的实体链接框架以及一种用于增强框架鲁棒性的负采样策略。在民航领域数据集上进行了对比实验，结果表明所提框架链接效果优于现有基准框架，并通过消融实验，验证了负采样策略的有效性。在负采样策略的作用下，该实体链接框架的Acc@top1高达0.875。

Entity linking method for civil aviation emergencies

Entity linking is the last step in building a knowledge graph, which is used to solve the problems of variation, ambiguity and unlinkable mention between entities. Relevant studies focused on the fields of medical care, biology and news, but there are few studies in the field of civil aviation. Therefore, this paper carried out a research on entity linking task in civil aviation domain, and found that there were many entity variants and less ambiguity in entity linking. To solve these problems, this paper proposed an entity linking framework based on semantic inference and a negative sampling strategy for enhancing the robustness of the framework. This paper conducted comparative experiments on the civil aviation field dataset, and the results show that the performance of the proposed framework is better than the existing benchmark frameworks, and the effectiveness of the negative sampling strategy is verified through ablation experiments. With the negative sampling strategy, the Acc@top1 of the proposed framework is as high as 0.875.