Perform sentiment analysis on the Amazon Fine Food Reviews dataset using TFLearn. Load data from an SQL database, construct a vocabulary processor, and test several deep learning models with variable parameters.

The Amazon Fine Food Reviews dataset is available on Kaggle and consists of 568,454 reviews of Amazon food items (link). The database contains the following information:

• Id
• ProductId - unique identifier for the product
• UserId - unqiue identifier for the user
• ProfileName
• HelpfulnessNumerator - number of users who found the review helpful
• HelpfulnessDenominator - number of users who indicated whether they found the review helpful
• Score - rating between 1 and 5
• Time - timestamp for the review
• Summary - brief summary of the review
• Text - text of the review

Scores and product reviews from the database are used to train a model that can predict if a review is negative, neutral, or positive based only the text contained in the review. The scores are given on a scale of 1 to 5. For this analysis the scores were grouped using the following convention: 1,2 = negative, 3 = neutral, and 4,5 = positive.

There are three base models available (contained in models.py): CNN, RNN w/ LSTM, and RNN w/ GRU. The loading of the dataset and vocabulary processing are handled in vocab.py.

The script is run with main.py and requires that a model be specified with '-m'. For example

$ python main.py -m cnn

The acceptable model names are cnn, rnn_lstm, or rnn_gru. The following optional arguments are also available:

• -b, batch size used for training: default=50
• -c, number of classes for softmax: default=3
• -d, whether to use dynamic rnn, bool: default=False
• -e, number of training epochs: default=5
• -f, minimum frequency for a word to be included in vocab processor: default=2
• -l, maximum length of text for vocab processor: default=200
• -m, name of model to use: options are 'cnn', 'rnn_gru', or 'rnn_lstm'
• -p, path to vocab processor model. Loaded if it already exists, created and saved if not: default=None
• -q, path to sql database: default='./database.sqlite'
• -s, number of samples to use for training: default=500000
• -u, number of hidden units: default=128
• -v, Tensorboard verbosity value, 0-3: default=0

Example: Dynamic RNN w/ GRU, save/load vocab processor as 'my_dictionary', batch_size of 100

$ python main.py -m rnn_gru -d True -b 100 -p my_dictionary