Text Understanding from Scratch Using Temporal Convolutional Networks

Astonishing? 3% better than bag of words after n days training on GPU's? I have misunderstood because I am not astonished.

What surprises me is that (BOW model + logistic regression) works just fine in most of the benchmarks(except for Amazon Review), interesting paper anyway. Could it be that because the vocabulary for BOW is limited to 5000, a lot of information is lost?

Open questions:<p>* Compare to RNNs with character level input.<p>* Compare to dedicated methods of sentiment analysis and topic categorization.

This paper is pretty fascinating, thanks! Having trouble visualizing it but getting there..