Experiment

(In point of accuracy)

• pre-trained embedding vs embedding pretrained > embedding (conv ,lstm)
• lstm(55) vs conv
  lstm(55) > conv
• mixed model(pre-emd&conv&lstm) vs each model (cnn ,lstm) mixed model > single model
• mixed model(pre-emd&conv&GRU(25)) vs mixed model(pre-emd&conv&lstm) mixed model(pre-emd&conv&GRU(25)) > mixed model(pre-emd&conv&lstm)

to do

• How to do better NLP preprocessing analysis
• autoML(DL)
• Improving performance ,what kind of methods there are in general.
• using Bert model