ctb result
HongyanJiao opened this issue · comments
HongyanJiao commented
I used default parameters and 8 layers bert ,trained on data ctb5.1.
I got 'FScore=90.50, CompleteMatch=28.82'
But the paper 'Cross-Domain Generalization of Neural Constituency Parsers'
refered F1=92.14, exactmatch=44.42
I was confused if there were anything wrong on my training, have you fine-tuned bert?
KnightCHB commented
Hey,I am using this model and ctb5.1 too.
Could I have your email? I want to ask you some questions. :-)
HongyanJiao commented
Sure,jiaohongyan0815@163.com
…---Original---
From: "KnightCHB"<notifications@github.com>
Date: Fri, Dec 27, 2019 19:18 PM
To: "nikitakit/self-attentive-parser"<self-attentive-parser@noreply.github.com>;
Cc: "Author"<author@noreply.github.com>;"HongyanJiao"<1412089534@qq.com>;
Subject: Re: [nikitakit/self-attentive-parser] ctb result (#45)
Hey,I am using this model and ctb5.1 too.
Could I have your email? I want to ask you some questions. :-)
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or unsubscribe.