Let us take the example for NES task.
Sentence 1: Vamsi loves Apple, especially the iphone.
Sentence 2: Vamsi loves Apple, it is healthy fruit.
So here for the model to detect Apple as a Company (in sentence 1) or a Fruit (in sentence 2).
As it needs information from both the sides we use BRNN.
So my doubt here is that, if we start from both the ends at t=0, then we reach the word Apple from left side RNN even before it comes from the right hand side RNN. So how is this problem tackled ?
Have you seen this ?