This is the Introduction of Natural Language Processing course in NTUST 2018. It uses Deep Learning to analyse the NLP problems. The result should be uploaded to Kaggle's competition, but since the game is not open to the public, I will not post the URL.
Introduce the base knowledge of Python, Deeplearning and TensorFlow.
Introduction to Python.
Introduction to TensorFlow : Install tensorflow & practice.
Introduce the concept of :
- Gradient descent
- Forward propagate
- Back propagation
- Easy Example by tensorflow
Practice the tensorflow.
Introduction to Keras and practice. The practice is to predict the probability of survival of Titanic passengers.
Introduce the Word Embeddings & Recurrent Neural Networks. Goal is the Document Classification used RNN: predict what class(catalogue) of article is obtained from PTT.
Introduce the Word Embeddings & Convolution Neural Network. Goal is the Document Classification used CNN: predict 'Push' and 'Down' number of an article. These articles are obtained from PTT.
Generative Adversarial Network is usually used to Image restoration.
We should create Generator and Discriminator to regenerate the white part of the picture with the painted white mouth. And GAN includes many kinds of models. I use the Pixel2Pixel architecture.
The more kinds of model realization can look here.