Skip to main content

New answers tagged

0 votes

How is the bidirectional context achieved in BERT?

In BERT, the context is considered bidirectional because there is no (causal) masking applied to the input sequence, allowing the transformer to process the entire sequence simultaneously. In contrast,...
frad's user avatar
  • 156
0 votes

How do transformer models handle negation in sentiment analysis

To answer this question, it makes sense to go over how multi-headed self attention works. Say we have 2 heads defined as follows (I am leaving small details out for simplicity): $$W_q^1, W_k^1, \text{ ...
Jayson Meribe's user avatar

Top 50 recent answers are included