cs224n Stanford deep nlp course

OS: Windows 10 64bit Please install missing libs if necessary.

Answers to CS224n problem sets, practice exams; only contains written and coding answers, original problem sets please refer to Stanford Official course site.
This is for personal interest, you are welcom to use the material but keep in mind this is not guaranteed to be the right answer.
It is also really appreciated if you could point out anything wrong or provide any suggestions on improvement of my answer.

Links:

• Course homepage - Orignial course materials
• Couse videos: - Free online course videos

Errors in PS:

1, PS1.2, in q2_neural.py function sanity_check, last parameter should be 'data', not 'params'.
gradcheck_naive(lambda params:
forward_backward_prop(data, labels, params, dimensions), data)

Contents:

language_models:

Play with different language models using Tensorflow/PyTorch
Experiment on multiple language models:

1. VanillaRNN/LSTM, to get my tensorflow work properly.
   The dataset is from @dennybritz's rnn tutorial @dennybritz's blog

2. Transformer
   Is from The Annotated Transformer

Assignment 1:

• Softmax
• NN basics (forward/backward propogation, Cross Entropy, Softmax and Sigmoid)
• word2vec (Cross Entropy loss, Negative sampling, Skipgram vs CBOW)
• Sentiment Analysis
  How to make it work under windows and python 3:
  1), Fix the data:
  The dataset in stanfordSentimentTreebank contains corrupted chars and has been fixed using windows notepad.
  First open the text, then save as another file in 'ANSI' encoding.
  2), Read file in byte mode:
  open(filename, "rb")

Assignment 2:

• Softmax tensorflow implement
• Neural Network Transition-Based Denendency Parsing
• Recurrent Neural Networks: Language Modelling

Assignment 3:

• Windows-Based NER
• RNN for NER
• RNN: GRU

Assignment 4:

Ongoing topics:

Word2Vec, CBOW, GloVe
RNN, LSTM, GRU
Dependency Parser, TreeRNN
Language Model,
Machine Translation,
Question Answering