Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

关于Extracting Multiple-Relations in One-Pass with Pre-Trained Transformers这篇论文 #16

Open
tc1713796338 opened this issue Sep 5, 2020 · 10 comments

Comments

@tc1713796338
Copy link

感谢作者分享的代码,对我的学习研究工作有很大的帮助。在readme最后的Paper realization中看到了Extracting Multiple-Relations in One-Pass with Pre-Trained Transformers这篇论文,请问在代码中哪里进行了改进,另外实体识别模型中,不知道我理解的对不对,为什么用了softmax分类而不是crf呢?

@yuanxiaosc
Copy link
Owner

yuanxiaosc commented Sep 5, 2020 via email

@tc1713796338
Copy link
Author

十分感谢您的解答

@tc1713796338
Copy link
Author

另外我看到在Main principle中,第二张图片给出了Attending to All Mention Pairs for Full Abstract Biological Relation Extraction论文的模型结构图,据我个人理解,是将实体识别和关系抽取的loss相加的过程用到了论文中的思想,不知道我理解的对不对

@tc1713796338
Copy link
Author

您好,实在多有打扰,您说的输出结果的方式用了论文中的方式,具体指的哪里呢,十分感谢,十分感谢

@tc1713796338
Copy link
Author

您指的输出结果的方式是说实体识别模块bert隐藏层输出用了论文中的方式嘛,我学的很吃力,所以向您请教,多有打扰,希望能得到您的回答,这对我的学习有很大的帮助

@yuanxiaosc
Copy link
Owner

@tc1713796338 你的理解是对的,该资源把bert当做特征抽取器,然后把上述论文当做输出层使用。平时工作很忙,来不及回复,见谅,学术交流可以发我邮件

@tc1713796338
Copy link
Author

十分感谢您的解答

@tc1713796338
Copy link
Author

看了您的解释对我理解论文大有帮助,只是反馈到代码上,又有一些迷茫,您说的bert当作特征提取层我很理解,Extracting Multiple-Relations in One-Pass with Pre-Trained Transformers这篇论文当做输出层,具体是在代码中如何体现的呢,是在下图中的实体识别模块中应用的吗,可是我没有找到论文中说的两个实体经过bert后,加和平均,然后再拼接的操作,我觉得还是我理解的不够透彻,您的回答对我有很大的帮助。再次感谢您前几次的帮忙和解答

@yuanxiaosc
Copy link
Owner

@tc1713796338 我回顾了下代码,发现只用到了 Joint entity recognition and relation extraction as a multi-head selection problem这篇论文的策略。有关这篇论文 Extracting Multiple-Relations in One-Pass with Pre-Trained Transformers你可以直接去搜他的官方实现看。我其实更建议你去搜一些顶会较新的论文来看,他们现在一般用pytorch写代码,看起来更容易。TensorFlow1.X的代码太恶心了,直接带劝退功能。

@tc1713796338
Copy link
Author

好的好的,十分感谢,我明白啦

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants