How to visualize the word embedding and word alignment? Are there any automatic methods?
Yeah, we have a script extract_embeddings.py in tools that let's you pull out the embeddings. There should be code in the forum to convert that to tensorboard format.
We don't have an alignment visualizer yet, but you can modify translation to print out the alignments.
I'll make it a request to document these two.
This http://nlp.seas.harvard.edu/2018/04/03/attention.html is a great resource too.
Most helpful comment
Yeah, we have a script
extract_embeddings.pyin tools that let's you pull out the embeddings. There should be code in the forum to convert that to tensorboard format.We don't have an alignment visualizer yet, but you can modify translation to print out the alignments.
I'll make it a request to document these two.