search

jinlanfu / NERmultimodal

Implementation of Adaptive Co-attention Network for Named Entity Recognition in Tweets in AAAI2018.
68 stars 9 forks source link

Accuracy issue #9

Open Saichethan opened 5 years ago

Saichethan commented 5 years ago

Hello, when I run the code I am getting scores as Image of scores

whereas In the paper it's given PER F1: 81.98 LOC F1: 78.95 ORG F1: 53.07 MISC F1: 34.02

Is there anything I am missing here? If that's because of incomplete data what are the scores you got for this data?

Saichethan commented 5 years ago

I did it 5 times, sure scores changed a little but not close to reported scores. where am I doing wrong?

Saichethan commented 5 years ago

25

On Wed, Sep 4, 2019, 9:05 AM thecharm notifications@github.com wrote:

how many epochs do you set?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/jlfu/NERmultimodal/issues/9?email_source=notifications&email_token=AIJNFOSVMM5KS7XE7C6VTD3QH4UIFA5CNFSM4HYYBKG2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD52HD5I#issuecomment-527725045, or mute the thread https://github.com/notifications/unsubscribe-auth/AIJNFOU6WPUWY3AFVWPK3RLQH4UIFANCNFSM4HYYBKGQ .

Amling2017 commented 4 years ago

Hello, when I run the code I am getting scores as Image of scores

whereas In the paper it's given PER F1: 81.98 LOC F1: 78.95 ORG F1: 53.07 MISC F1: 34.02

Is there anything I am missing here? If that's because of incomplete data what are the scores you got for this data?

你是按照原文的环境配置，复现完成的吗？

Saichethan commented 4 years ago

@Amling2017 no even after parameter tuning we were able to get at max 67.5 which is slightly better than ma hovy at al.

Amling2017 commented 4 years ago

@Amling2017 no even after parameter tuning we were able to get at max 67.5 which is slightly better than ma hovy at al.

你的代码上传到GitHub了吗？方便交流一下吗？