This is my current project for finding the sentiment for Reddit posts and comments and tweets on Twitter. The Reddit search utilizes a post or comments upvotes and downvotes to add a “scaler” to that post or comments positive or negative sentiment score. I’ve recently integrated Twitter into my program and plan to research if a tweets “likes” and retweets would be suitable to give a tweets positive or negative sentiment more “weight”.
I built my own sentiment module using NLTK, which is imported into my program and that module is what does the sentiment analysis. The program runs via command line and as a web app using Flask. All results are stored in a database.
The intent of this project was to provide a business a way to dissect reviews from non-traditional rating websites such as Twitter. For example, when someone posts a review on Yelp for a restaurant they’re able to leave a numeric rating (4/5 stars for example) and have the option to describe what they did and didn’t like about their experience. On websites such as Twitter this rating system isn’t a feature of the website, but a rating can be derived by reviewing the sentiment of individual tweets and the overall sentiment of the tweets parsed.
Besides giving a business a rating, the sentiment analysis can also use word or phrase frequencies to determine what someone or groups of people liked or disliked about their experiences without have to read through all individual reviews.