what you did for embedder?
I meant that fastai preprocessing and ability to finetune are two important bits that might be hard to replicate, and it might be easier to add NER to fastai.
Do you know if there’s anyone working on this? I tried going on from @sebastianruder’s branch , but I’m having a hard time with it. I’m not too experienced developing fastai code and framework.
I wasn’t paying attention to NER and ulmfit as i’m focusing on closing the classification. I’m thinking about taking on NER in Feb / March, once classification is done. If you need something that works, try zalando flair they have pretrained networks for NER.
1 Like
I see. I’m writing a survey on NER, and was expecting to compare all these different representations, such as ELMo, Transformer, Bert, ULMFiT, Flair, etc. The only one I wasn’t able to evaluate so far was ULMFiT.
1 Like
Please post a link to your survey, if possible. I am also looking into implementing NER with ULMFiT, but it does not appear to be as straight forward as originally stated in the ULMFiT paper.
2 Likes
Hello @davidpfahler, where you able to make progress implementing NER with ULMFiT?. Please let know.