Bait Bully

Full stack of Bait Bully

Inspiration

What inspired us to create this application was our shared annoyance toward fake news and clickbait articles which ultimately get in the way of our desires of accurate and unbiased news and information. However, as humans, it is hard to consistently determine what articles are or are not trying to deceive us in favor of their own ad revenue. Thus, we were inspired to utilize the consistency and accuracy brought forth by Google Cloud's AutoML machine learning tool, as well as the ease of use and serverless cloud services such as AWS Lambda in order to create an application that is constantly and efficiently running when necessary.

What it does

At the highest level, Bait Bully is a chrome extension which swiftly identifies if an article on a Google News page is clickbait or not, based on the title of the article. These clickbait predictions are based on millions of data pre-determined by an acknowledge data set.

How we built it

Various services and methodologies were implemented in the creation of this application. While perhaps simple sounding at a high level, the application travels through a complex stack in order to ensure the durability of the application. First, using Chrome API, we created the extension, which would be responsible for gathering data from the Google search webpages and then passing it to the AWS Lambda function. This function was then responsible for constantly listening for requests from the extension, and then processing these requests with a Google Cloud's AutoML NLP model. This model was trained using millions of data that was publicly available for the purpose of NLP model creation. Lastly, these classifications are passed back to the Lambda function, back to the extension, which in turn modifies the webpage.

Challenges we ran into

Our team encountered many challenges in the creation of our application. Many of these challenges surfaced as a result of our wide use of new and unfamiliar technologies to create what we thought would be an ambitious goal. For example, we encountered challenges with regards to authentication errors when using Google Cloud, hours of training models on the AutoML platform, and hours of planning in order to determine a suitable stack that runs almost exclusively in the cloud. Furthermore, many other errors surfaced throughout the process mostly as a result of our varying specialties, however through healthy communication we where able to solve the majority of our problems.

Accomplishments that we're proud of

What we are most proud of is that we were ultimately able to implement a product that we thought might be too ambitious in the 24 hour time frame. Primarily, we where able to apply a whole suite of new technologies and methodologies that are currently being implemented in the real-world to our application, with almost no prior knowledge to these topics before the hackathon.

What we learned

While our team learned much with regards to technologies such as Github, AWS, Google Cloud, machine learning, and much more, what we learned the most was how to work as a team. More specifically, we all learned that effective communication is vital in a team setting, where it is necessary that everyone's individual skills is utilized to the maximum. This was especially applicable in our case considering our team consisted of an EE, CS, BME, and Chemistry major.

What's next for Bait Bully

While we are very satisfied with how Bait Bully turned out, there is definitely room for generalizing our application to more cases. Such cases include support for other browsers, other webpages, and perhaps even email. Furthermore, Bait Bully has the possibility of manifesting into a publicly available extension.