Kevin Leong
Computer Science (B.S.) & Creative Computing (B.A.)
Sentiment Analysis
In this project, I create a program to analyze
the sentiment of movie reviews. I used 40,000 movie reviews labeled positive or negative in sentiment as an initial data set. Then, I used 10,000 movie reviews to test the accuracy of
my algorithm.
In order to create this sentiment analyzer, I parsed in 40,000 movie reviews and assigned a positive or negative point value to all words from each review. The words I parsed were added to a map in which the key was the word and the value was the point value of the word. Each word got a positive point for being in a positive review and a negative point for being in a negative review. Next, I removed all neutral stop words from the map in order to get a more accurate measurement of sentiment. This gave me a map of emotional words and a point value for each word.
In order to test the algorithm, I took 10,000 movie reviews and parsed them word for word. Then, using the map of word values, I added up the point values for each word in the review. This resulted in each review having a total review score of being either positive or negative. I used this score to assign each review a sentiment. I compared the sentiment that I assigned to the review to its actual sentiment in order to get an accuracy score. This algorithm resulted in an accuracy of 68.61%.
In this project, I also implemented a custom string wrapper class.
Click here for the full Github Repository
Image Source