Is the attention explanation of why the model tells like this? I've seen that there are many discussions about this. (Image attention visualizations were not that good I think)