ModBuddy is a reddit moderator tool that can detect and flag harassing and vulgar comments. It was inspired by the Hack Harassment challenge at QHacks 2017

ModBuddy is composed of two parts- a RESTful API that performs analysis on text, and a chrome extension that submits comments from Reddit for analysis.

ModBuddy uses indico.io APIs to run sentiment analysis and a custom anti-harassment text model. ModBuddy also includes detection for many common vulgar words.

The chrome extension makes POST requests to /analyze, and based on the response will highlight comments to indicate a potentially harassing or vulgar comment to the moderator.

• Go to chrome:://extensions, enable developer mode.
• Click Load unpacked extension
• Navigate to and select the ModBuddy root directory

Here's an example of how harassing comment highlighting works

• Our MVP was to highlight comments, but we also wanted to have a 'moderator menu' that allows bulk navigation to, selection, and deletion of comments. However, we were unable to complete the deletion feature in time for the deadline.
• A future item would include allowing custom rules. Right now, the chrome extension flags a comment if any of the returned data parameters (such as containsBadWords) is out of spec

The ModBuddy server is built on the flask framework. You must have a valid installation of Python 2.7 and install the following packages using pip

• flask
• indicoio

After installing prerequisites, you must create a file called indico-api-key.txt in the flask/static folder, containing only your API key from the indico dashboard. To get an API key, you can sign up for a free account. Then, run python application.py, and the server will be available at 127.0.0.1 on port 80.

To demonstrate the functionality without an API client, a simple web form was created. To train the model, navigate to /train and enter comments (one per line) as "non harassing" or "harassing" and click Submit Training Form. Then, navigate to /analyze (root will redirect you there), and enter a comment to be analyzed. The response will show the JSON objects returned by the API.

Example of an acceptable comment

Example of a harassing comment

/analyze

/train

To deploy on AWS elastic beanstalk, create a Python 2.7 instance in your desired configuration (Low cost instances are eligible for the free tier). Create a zip file with the files found in the flask folder. (application.py should be at the root!). Then, simply upload the zip file as the application you wish to deploy.

Currently, only the English language is supported. With a small amount of training data, classification may not be as accurate as possible.