Repository navigation
farasa segmentation #174
Answered
by
WissamAntoun
yassmine-lam
asked this question in
Q&A
farasa segmentation
#174
|
Hi, Based on your experience, which model is better for Named entity recognition and why: a BERT model with or without farasa segmentation. Does segmentation affect sequence labeling tasks? Many thanks |
Answered by
WissamAntoun
Oct 9, 2022
Replies: 2 comments
|
for NER i would go for a BERT without segmentation. in our paper we experimented with both, and found out that segmentation was hurting performance. We guessed that it is caused by the label being put on the "ال" (or other prefixes) at the beginning of most words. |
0 replies
Answer selected by
yassmine-lam
|
Ok. Thank u very much for ur answer |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
for NER i would go for a BERT without segmentation.
in our paper we experimented with both, and found out that segmentation was hurting performance. We guessed that it is caused by the label being put on the "ال" (or other prefixes) at the beginning of most words.