Named entity extraction does not work with few examples. It does not recognise the entities that have not been provided in the training data file. But it works when provided with at least 2500 examples. Then, it recognises entities that have not been provided in the training data file. I am using tensorflow embedding.Can anyone suggest any reason behind this?
Hi @anubhavpnp, this is to be expected – with only a few examples, the model can’t be expected to transfer learn to unseen examples that well (what do you mean by “a few”)? With lots of examples, it should pick up on new unseen values for entities (not new entities though, those have to be defined in your domain file). I don’t think it should require 2500 examples for that though! But indeed, more training data always helps 
hi @erohmensing…Thanks for the reply. By few , I meant 10-15 examples.Can you please point me to a basic tutorial on why this should be the case. What changes when the number of examples get large enough?I am a newbie to machine learning. Thanks for bearing with me.