Construction of Entity and Entity Relationship Annotation Corpus for Test Identification. Issue 1 (1st November 2022)

• Record Type: Journal Article
• Authors: Wang, Bo; Xu, Zixi; Huang, Jianmei; Sun, Teng; Mu, Xiaofeng
• Is Part Of: Journal of physics. Volume 2363: Issue 1(2022)
• Publication Date: 2022-11-01
• Subjects: Test identification -- entity -- entity relations -- corpus construction; Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/2363/1/012018
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract : Currently, the test identification corpus faces the problems of small scale, poor consistency and poor standardization, which is difficult to meet the needs of intelligent applications such as knowledge mapping, question answering system and knowledge reasoning. Based on the entity labeling system for test identification developed by CCKS evaluation conference, this paper developed and improved the labeling specification. On this basis, the original text of test identification was obtained through news, Wikipedia and other channels. The entities in the original text were manually labeled and collated by the annotator, and the entity annotation corpus for test identification was constructed. The corpus constructed in this paper has a total of 86, 000 words, with a total of 5980 entities and 4982 entity relations. The consistency of multi-round annotation is 86% and 85%, respectively.