DeepDarkHomework

NLU and SER project in Deep Learning Class.

##Todo

• Read file
• Write file
• Evaluation (Maybe not needed.)

NLU

Natural Language Understanding python3 snips_nlu_t1.py

• Stop confuse
• Capsule NN (?) https://arxiv.org/pdf/1812.09471.pdf
• Attention based RNN (?) https://arxiv.org/pdf/1609.01454.pdf
• CNN based TRIANGULAR CRF (?) http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.642.7548&rep=rep1&type=pdf
• Slot-Gated Modeling (?) https://www.aclweb.org/anthology/N18-2118.pdf
• RNN with a max-pooling layer (??) https://www.ijcai.org/Proceedings/16/Papers/425.pdf
• Pretrained model

SER

Speech Emotion Recognition python3 ser_crnn.py.py

• FFT or z-transform
• Mel filter
• Preprocessed inputs are provided (logMel features) (MFCC)
• Padding of features.
• CRNN