140609 Xi Ma
1.Using a new method ,which is based on comparing the cross-entropy,according to domain-specfic nad non-domain-specfic language model,for each sentence of the text source used to produce the latter language model.The non-domain text source is baiduzhidao and weibo.
2.Supplement the weight of each word in the new vovabulay.
1.Continue to extract sentences using the method based on comparing the cross-entropy.
2.Test ppl of the language model using different training set to train.