This porject is porting our original code in Theano to Tensorflow. Still in very early stage and under heavy development.
Implementation of Neural Language Correction (http://arxiv.org/abs/1603.09727) on Tensorflow
Tensorflow 0.9
To train character level model (default):
$ python train.py
To train word level model:
$ python train.py --tokenizer WORD
# 尝试科大VPN之后还是socket error
# 目前可深究小问题:
# nlc_data 中 正则表达式处理部分未细究可查看py的正则并深究并作下正则解析过程记录
#主要工作范围缩小至 nlc_model 类定义, nlc_data 的正则处理
$ python decode.py