Spam Slayer

Spam Slayer Web Page

Inspiration

User review is one of the most important factors that customers want to use when they shop online. However, the authenticity of the reviews is not guaranteed. Some of the reviews are generated by machine. The group members used Amazon a lot. It turns out that humans hardly succeed when we try to distinguish between truthful and computer generated reviews (with a 40% accuracy). Then can machine learn it?

What it does

This application accepts the URL for an Amazon product and filters out the reviews it suspects to be fake/computer generated. A new rating for the product is also generated.

How we built it

The website is built with dJango while the amazon reviews were crawled using scrapy. The reviews are analyzed by a CNN model trained on over 20k Amazon reviews.

Challenges I ran into

We tried to integrated many features in a short amount of time and we unable to perfect each feature as desired. We especially had problems with using scrapy as it was not performing as expected from its documentation. Further, getting dJango to perform fast enough for practical use was a big issue.