Closed Brook4846 closed 6 years ago
The answer module in paper is GRU (RNN), so you can acquire more than a one word. But my implementation acquires only one word.
Well that's reassuring. Would you be kind enough to send me links of these implementations on Github if there are any you're aware of.
Sorry, I don't know it.
Well, thanks anyway...
I've went through the DMN paper and have seen some of the examples they presented for the question answering implementation using the bAbi dataset and wanted to know weather we can acquire more than a one word answer before trying out this implementation.