Closed xinghai-sun closed 6 years ago
@cxwangyi @kuke @xinghai-sun Hi, as mentioned in the paper, a language model has to be trained to improve the generating results and the LM is a critical component to ensure the performance. The language model is trained on texts crawled from commoncrawl.org using KenLM toolkit. However, we need more details to train such a language model. Any possible to get the trained language model or text dataset trained on?
- 英文的语料应该很多,不一定拘泥于paper提到的语料,我们可以先小语料试,例如PTB。
- n-gram LM训练的工具尽可能先用KenLM;如果不用,也尽可能保证 smooth的方法对齐或合理。
- 重点关注model loading和inference的接口设计,并且做好和beam search decoder的联调。
- 联系NLP或者SVAIL看看有没有现成的powerful LM model,中英文都问下,请@lcy-seso 协助下。
@lcy-seso 请问有现成可用的LM model可以用吗