中文

症状词比喻用法检测以改善个人健康提及检测

计算与语言 2019-07-05 v2 信息检索

摘要

个人健康提及检测旨在预测给定句子是否是对健康状况的报告。已有工作指出,当症状词(即关注症状的名称)以比喻意义使用时,该预测会出现错误。因此,我们将最先进的比喻用法检测与基于 CNN 的个人健康提及检测相结合。为此,我们提出两种方法:一种基于流水线的方法与一种基于特征增强的方法。在基于特征增强的方法中,引入比喻用法检测使个人健康提及检测的 F 值平均提升 2.21%。本文展示了利用比喻用法检测来改善个人健康提及检测的前景。

关键词

引用

@article{arxiv.1906.05466,
  title  = {Figurative Usage Detection of Symptom Words to Improve Personal Health Mention Detection},
  author = {Adith Iyer and Aditya Joshi and Sarvnaz Karimi and Ross Sparks and Cecile Paris},
  journal= {arXiv preprint arXiv:1906.05466},
  year   = {2019}
}

备注

To appear at the 57th Annual Meeting of the Association for Computational Linguistics (ACL 2019) (The second version updates the name of a cited paper. A detailed note from the cited author is here : https://github.com/commonsense/conceptnet5/wiki/Citation-complications )