Sentiment Analysis
Dataset: Stanford Sentiment Dataset and IMDB.
Goal: apply different classifier for sentiment detection on two commonly-used datasets above. Classifiers used: (1) Classic classifiers: inear classifier, logistic classifier, perceptron and SVM; (2) Neural Network: single-layer feedforward neural network; RNN (with basic, GRU and LSTM units).
Code Structure: Here is a list of source scripts: classifiers.py: implement simple classifiers like linear classifier, logistic classifier, perceptron and SVM. data_util.py: data preprocessing, feature dumping and loading. fnn_tf.py: implement single-layer feedforward neural network for sentiment classification. rnn_tf: implement recurrent neural network with basic, GRU and LSTM units. vocab.py: implement class Vocab for embedding processing. word2vec_tf.py: implement functions for word embedding training and saving. IMDB_src/parser.py: parse the raw data in IMDB review dataset. IMDB_src/vocab.py: transforms IMDB text into embedding.