Inspiration

Too many junk comments on the Internet

What it does

Detects the spam comments on youtube and masks them

How I built it

We pulled the training and testing data from UCR. We cleaned the data with a python pipeline with lowering alphabetical letters, stripping the HTML and email addresses, congregate numbers, detected dollar signs.

We used a support vector machine with Gaussian kernel. The final accuracy we get is 86.48%.

Challenges I ran into

Figured out the meaningful features. Cleaned up the data with regular expression.

Accomplishments that I'm proud of

Finished building the entire data processing pipeline and model production.

What I learned

Collaboration in between a full stack project.

What's next for direHusky

Finish up the chrome extension part. Publish online. Potentially optimize the model to have a better accuracy. Can also be applied to other Internet scenarios.

Share this project:

Updates