Movie Review Sentiment Analysis

Home
Positive Movie Review
Negative Movie Review

Inspiration

The job of a film critic is a highly responsible job. These days, people have become smart enough to read or watch a movie review investing their money in a ticket. The lot depends on the movie reviews nowadays and therefore they are important.

Films reviews provide advance information about the film before it reaches the final audience. Generally, the critics see the film before the actual release or on the day of release and they review it based on what they feel. A good film critic talks about various aspects of film making like cinematography, acting, sound design, music, etc. without giving any spoilers so that audience can decide whether they want to spend money on a film or not.

Movie reviews help in understanding films better

Apart from giving information about the film, the reviews are important because they make audience understand the meaning and connotation of the movie, and sometimes the story. Nowadays, films with open endings or complex story lines are made and it is not always easy for the audience to decode the movie and hence they turn to their favorite critic for additional information. For example, the movie Mother starring Jennifer Lawrence has so many underlying themes and metaphorical meaning which are difficult to understand

What it does

This app has been built to predict the sentiment of a movie review whether is good or bad. If the code returns 1% above its prints out Positive Review (Movie is Good) which literally means the movie comment or review signifies that the movie might be good for other audience. Also, if the code returns less than 1% its prints out Negative Review (Movie is Bad) which means the movie comment or review signifies that the movie might be bad for other audience.

Model Used

Logistic Regression Analysis Logistic regression is a statistical analysis method to predict a binary outcome, such as yes or no, based on prior observations of a data set. A logistic regression model predicts a dependent data variable by analyzing the relationship between one or more existing independent variables.

In this case if our model predicts 1 it means the movie review is positive, making the movie a good one, if model predicts less than 1 it means the movie review is negative making the movie a bad one.

We used CountVectorizer to transform a given text into a vector on the basis of the frequency (count) of each word that occurs in the entire text. In order to get most occurred text we used TfidfVectorizer in combination with Logistic Regression

Accomplishments that we're proud of

Using Long short-term memory we were able to get a test accuracy of 0.8750 and a loss of 0.4559

What we learned

I got to learn a lot about CountVectorizer and I was able to build a web application can source data from the IMDB which can be accessed via http://movie-sentiment.rogerkoranteng.com

What's next for Movie Review Sentiment Analysis

In order to improve model efficiency, I will be building an API to automatically serve our application with live comments from the IMDB for easy update and training