seq2seq
seq2seq copied to clipboard
Examples for input with multiple features??
Hi,
After implementing the NMT tutorial on the Toy Data, I succesfully implemented a similar approach for character level seq2seq model. In the tutorial, inputs are single characters (from an array of characters) but I was wondering, how we are supposed to give inputs if we have 2 features for the same character.
For example- for the word "CAT", inputs according to the tutorial will be "C" "A" "T". But what if I want to give one more feature along with the characters?? Suppose I also want to give POS tag of "CAT", then what is should be the format of the input??
@vikas95
Did you manage to find a way for this in the mean time? I was wondering if a 'naive' approach (like in this paper with the mixed encoder: https://arxiv.org/abs/1705.01020 ) would work well.
I wrote a blog post addressing this feature - https://iamsiva11.github.io/extra-features-seq2seq/. Hope you find it useful.