基于COAE2016数据集的中文实体关系抽取算法研究

实体关系抽取是知识图谱技术的重要环节之一。英文实体关系抽取的研究已经比较成熟,相比之下,中文实体关系抽取的发展却并不理想。由于相关语料的匮乏,中文实体关系抽取的发展受到了一定的限制。针对这一问题,COAE2016在任务三中提出了中文实体关系抽取任务。通过分别使用了基于模板、基于SVM与基于CNN的实体关系抽取算法解决了这一问题,并根据其在COAE2016任务三的评测数据集上的效果,对比分析了三种实体关系抽取算法的优缺点。实验证明,基于SVM的算法和基于CNN的算法均在评测数据集上表现出了良好的效果。

Chinese entity relation extraction algorithms based on COAE2016 datasets

Entity relation extraction is one of the important procedures of knowledge graph technology. Research on entity relation extraction in English is comparatively developed. By contrast, the development of Chinese entity relation extraction is not ideal, and it is mainly because the lack of corpus. In order to solve this problem, COAE2016 proposes a Chinese entity relation extraction task in task 3. In this paper, we use three algorithms to solve the problem: a pattern based algorithm, a SVM based algorithm and a CNN based algorithm respectively. Then, we analyze the advantages and the disadvantages of the three algorithms according to the effects of the dataset in COAE2016 Experiments show that the SVM based algorithm and the CNN based algorithm are useful to extract entity relation.