3 ms·
And bases the argument on a single feed forward network, nothing about RNNs which most good NN NLP results are based on.
by sprobertson 10y ago
And bases the argument on a single feed forward network, nothing about RNNs which most good NN NLP results are based on.
- habeanf 10y agoGoogle's SyntaxNet (as of Andor et al 2016) achieves state-of-the-art accuracy for dependency parsing without RNNs.