Unclear results for Multi-Candidate Word Segmenation using Bi-directional LSTM Neural Networks

[p16i]

Hi,

I've found the paper "Multi-Candidate Word Segmenation using Bi-directional LSTM Neural Networks" for your repository. While the approach in the paper sounds interesting, the details in mentioned in the experiment setup sounds very strange to me.

As you're one of the authors, I wonder how the algorithm would perform if we try it on the whole test set. Obviously, the numbers would look differently. Would you have time to do that and update the plot and the numbers in the table accordingly?

Thank you very much in advance for your clarification.

The figure is taken from https://drive.google.com/file/d/1x8JmqQFlbMev0fiqCx7wM5YU-twxHmTu/view.