Movie Review Sentiment Analysis
--------------------------------------------------------------
This is a guide on how to run our Sentiment Analysis program
Brief overview:
Our program is run on the IMDB movie review data set that
is found in./data. The directory contains reviews for
training purposes as well as testing purposes.
------------------------------------------------------------
Running the program:
To run the program, execute python3 main.py
------------------------------------------------------------
Output:
Performance metric with different extraction features
------------------------------------------------------------
Changing Parameters:
- To change the extraction feature, in mt.py,
choose between
bowFm() (bag of words feature matrix),
tdIdfFm() (tf-idf feature_matrix) or
normalizedWfFm() (normalized word frequency feature matrix) in
getSplitData() on lines 260 and 261 in in mt.py when passing in
the training and testing data
- To remove or keep stop words, in main.py,
Comment out a section if-statement on line 116 in mt.py
Change the line in the method getDictNoSw() to say this:
if word not in word_dict : #and word not in stop_words:
- To remove proper nouns from the review,
use getDataNoCaps() instead of getData()
in main.py. Simply comment out the four lines of code
which use getData from lines 7 to 10 or lines 14 to 17
- To run quadratic kernel, on line 56, change the setting
to "poly" instead of linear, change Degree to 2, and add(and an r value coef0=1)
- To change the number of reviews the experiment
is run on changes REVIEWS in mt.py on line 17
REVIEWS represent the number of positive and negative reviews to be chosen.
REVIEWS = 2500 means, running the experiment on 5000 movie reviews.
suvinay17/MovieReviewSentimentAnalysis
Binary classification of Movie Reviews with SVMs. Uses NLP techniques like TF-IDF, Bag of Words, Removing Stop Words, and Normalized Word Frequency.