Why is this image not working?

Inspiration

For ArtiFact, we wanted to help others recognize when media was using first impressions of their content to try to get users to view the rest. It’s a problem everyone has faced at some point; you read a flashy headline and are enticed into viewing the content before ultimately realizing that the actual content was a massive disappointment. Commonly known as “clickbait,” these tactics are used in the media to draw attention and “clicks,” ultimately resulting in a more successful career in media for the creator. Seeing as this is such a common annoyance in our everyday lives nowadays, we wanted to create a solution to this problem. So, to counteract this recent trend, we’ve used our knowledge of machine learning and web/app development to create ArtiFact.

What it does

As previously mentioned, ArtiFact is an Artificial Neural Network taking on the form of an application or a chrome extension that uses machine learning to determine how likely it is that a headline is misleading a user on a scale of 1-100. The test cases for the neural network were all stored on CockroachDB for various reasons later discussed in the devpost. The chrome extension of the application will also show a warning for news sites that have notoriously poor-performing average ArtiFact scores by scanning through the previously rated headlines stored in a database using CockroachDB. This application, however, does not determine the rating based on the actual content of the article. Only the headline is fed into the neural network, so there will be cases where an article’s content lives up to the headline given, and this will have to be up to the user whether or not they would like to engage in that content. Another shortcoming of ArtiFact is that it likely will not always be 100% correct because it is a neural network. This is a flaw that we will continue to attempt to chip away at by feeding the neural network more data whenever necessary.

How we built it

When creating the application and the neural network, we used python 3.9 since we were the most comfortable with it for machine learning and graphical user interfaces. We used HTML, CSS, and JavaScript for the chrome extension to create the extension and include all the necessary features. We also used CockroachSQL when sending and receiving queries between the database and the script.

Why is this image not working?

CockRoachDB Implementation

As for the usage of CockroachDB in this project, as stated before, we implemented CockroachDB to store our neural network’s test cases and easily add more consistently. The first upside of using CockroachDB, as mentioned in the video, is its serialized isolation capabilities. This practically eliminates the possibility of any dirty reads or writes, perfect for our use case when the test data must be excellent to teach the AI correctly. This is also helped by its atomicity (the requirement of transactions to be all or nothing), meaning partial transactions can’t happen. Not only that, CockroachDB allows us to store any headlines previously submitted by the users to create a collection of ratings. This collection is then used to flag certain websites as highly unreliable. Finally, the best feature of CockroachDB we’ve found was the advanced admin tools that allowed us to visualize and view the data and traffic in a super organized and understandable manner.

Business Viability

The Business Viability of this application comes from the fact that a) It solves a reasonably common annoyance in people’s lives and trains them to look for content differently, and b) it is an expandable idea that can be used for a multitude of other purposes. For example, for students who don’t want to waste time looking through articles with misleading headlines that may be totally exaggerated or emotionally charged, ArtiFact can come in handy to scan through web pages to search out appropriate headlines. The web application and extension can be easily monetized with targeted advertisements and anonymous usage data collection. Upkeep costs are low as the service is run entirely on cloud resources; thus, the cost and service provided can be easily adjusted based on customer demand changes.

Challenges we ran into

For challenges, our main one was web development as none of us are particularly too experienced in it, so the chrome extension took some learning to figure out. Also, there would be times during the artificial intelligence testing where we’d run into a dead network, so we would have to work our way backward to find out where it came from.

Accomplishments that we're proud of

This was our first time implementing a neural network into an application, so we were super glad to add all the data in and have it work for the first time. Also, implementing the chrome extension for convenience was a step we believe in the right direction for this type of application.

What's next for ArtiFact

Mainly discussed in the Business Viability section, but we are planning on adding even more cases as time goes on and maybe even a way for users to submit data themselves for the neural network to learn from. ArtiFact is also going to continue exploring new ways to improve and optimize the AI network going forward. Our end goal for ArtiFact would be content-aware detection, comparing what is said in the headline to what is actually written in the article.

Share this project:

Updates