Hi Akshat, thanks for the post. What if we represent each word of the intent with its Word2Vec(assume vector of size 50) and for hence sentence(assume we have 3 words), so sentence can be represented as sum of three vectors(size 50) divided by 3. So sentence will be a vector of size 50(instead of maxlength in your case). And next structure will be same. Does it make sense ? Please explain(I am not getting good accuracy on my dataset with what you have done ).

Thanks in advance !