Sentiment Analysis of Movie Reviews

Packages Used

There is no additional installation needed to run this program.

The modules and their versions used for this project are as follows:

• pandas (Version: 1.4.2)
• nltk (Version: 3.7)
  • submodule: vader
  • submodule: corpus
  • submodule: stem
  • submodule: tokenize
• sklearn (Version: 1.0.2)
  • submodule: feature_extraction
• numpy (Version: 1.21.5)
• matplotlib (Version: 3.5.1)
• seaborn (Version: 0.11.2)

Running the program

You need to be at the root directory sentiment_analysis of the project to run the program. There are a few options that can be provided with the run command.

You can run this program with the reuqired arguments as follows:

python NB_sentiment_analyser.py <path to train file> <path to dev file> <path to test file>

The optional arguments are:

• -classes: number of classes -> default 5 classes (options: 3,5)
• -features: features -> default is considering all words (options: all_words, features_word_type, features_tfidf)
• -output_files: option to display output files -> default is producing no output files
• -confusion_matrix: option to display the confusion matrix -> default is showing no confusion matrix

Since our train/dev/test datasets are in a directory called moviereviews, we will need to specify that when we are supplying the names of the input files. An example run command where we want to consider 3 classes with producing output files and confusion matrices for each class will be:

python NB_sentiment_analyser.py moviereviews/train.tsv moviereviews/dev.tsv moviereviews/test.tsv -classes 3 -output_files -confusion_matrix

If output_files argument is provided in the run command, the produced output files will be created in the predictions directory.