implement the paper" Very Deep Convolutional Networks for Natural Language Processing"(https://arxiv.org/abs/1606.01781)in tensorflow,just 9 layers(without shortcut).I change the last k-max pooling layer into max-pooling.
Parts of code is based on https://github.com/amygdala/tensorflow-workshop/tree/master/workshop_sections/cnn_text_classification ,which is based on the https://github.com/dennybritz/cnn-text-classification-tf,parts is based on https://github.com/scharmchi/char-level-cnn-tf
The data of the experiment is dbpedia, accuracy of the experiment is 0.9782,while the paper reports 0.9865