“NLP Status Report 2017-7-10”版本间的差异

• reproduced the couplet model using moses

• continue to modify the couplet

• Tried a seq2seq with style code model but it didn't work.
• Coded attention-based seq2seq NMT in shallow fusion with a language model.

• Complete coding and have a try.
• Find more monolingual corpus and upgrade the model.

• run two versions of the code on small data sets (Chinese-English) and tested these checkpoint

    found version 1.0 save time about 0.03s  per step, 
    and these two version  has  similar complexity and bleu values 
    found that the bleu is still good when the model is over fitting .
    (reason: the test set and the train set of small data set are similar in content and style) 

• run two versions of the code on big data sets (Chinese-English) . OOM（Out Of Memory）

    error occurred when version 0.1 was trained using large data set，but version 1.0 worked 
    reason: improper distribution of resources by the tensorflow0.1 frame leads to exhaustion of memory resources 
    I had tried 4 times （just enter the same command）, and version 0.1 worked 
    found version 1.0 save time about 0.06s  per step, and these two version  has  similar complexity and bleu values 

• downloaded the wmt2014 data set ,used the English-French data set to run the code and

   found the translation is not good (reason:improper word segmentation)

• do word segmentation on wmt2014 data set
• run two versions of the code on wmt2014 data set
• record the result and do analysis
• learn and train moses(use big data sets (Chinese-English))