Shortcut for tagging unseen/unlabeled data about bert-e2e-absa HOT 4 OPEN

I am also interested in this, as far as I can see the method only works with input data that has labels (the part after ####). Did you find a way to do this?

@lixin4ever Is there a way to input unlabeled data?

from bert-e2e-absa.

For the part after ####, they are just placeholder and not used during inference (i.e., the prediction on your own data) and you can set an arbitrary valid tag (e.g., O, B-POS, I-NEG, E-NEU, and so on) for each word to facilitate the running. Note that you should keep the format identical to that in the provided data files (see the data files in the folder ./data).

from bert-e2e-absa.

Thank you, I adapted your comment and it works.

from bert-e2e-absa.

Hi, We want to train your model using our own dataset. We are facing some difficulty to label our data
one example is given here
Guide us to label the dataset
has the worse customer service I was on hold for over an hour and when I finally got though the agent couldn't be bothered to help me. She put me back on hold then hung up! ####has=O the=O worse=O customer=T-NEG service=T-NEG I=O was=O on=O hold=O for=O over=O an=O hour=O and=O when=O I=O finally=O got=O though=O the=O agent=T-NEG couldn't=O be=O bothered=O to=O help=O me.=O She=O put=O me=O back=O on=O hold=O then=O hung=O up=O !=O

from bert-e2e-absa.