10_000 samples & 10_000 labels

Hey Community, I have a data set which each sample has it’s own label, for instance :
I have 10000 sample which each sample has one word as label, and each label is unique for that sentence, this made 10000 training samples with 10000 labels.
Is anyone here has an idea about how to do this or a toy code?

Thank you so much for your help

I don’t think that 1 single sample for 1 label is going to learn well.

But, in general, assign a label to a sinlge word is in the perimeter of the Named Entity Recognition (NER). You can find many tutorial with Keras/TF like:

@Bhack it’s not NER, anyway thank you so much for your reply.

I don’t know exactly what your is your domain but the problem setup It seems to be similar to NER.

With just 1 sample per class it seems similar to one-shot NER.

So the most similar thing that I can think about your data is finding some idea in few-shot NER approaches:

https://arxiv.org/abs/2012.14978

https://arxiv.org/abs/2105.07464

1 Like

@Bhack thank you so much