As a Resident Advisor (RA) on North Campus, I've been finding ways to build a community on my floor. We had a suggestion box in the hallway for residents to provide their suggestions for the community. Unfortunately, no one ever puts anything there, either because they didn't want to or someone took them away. However, when asked about their suggestions, residents always have a lot to say. Where's the problem?
By talking with residents, we learned that what people needed was a friendly transparent platform where ideas can be exchanged rather than collected. Our confession wall is such a platform that helps people anonymously post any suggestings or feelings they have in mind about the community without having to worry about being recognized or overlooked.
What it does
Big Red Confession Wall is a safe and dynamic web application where any member of the community can write anonymous posts and like/dislike and comments on others' posts. Before, suggestions for a community were usually collected in a top-down manner. For example, leaders send out surveys to community members. There is a problem with this process because everyone's ideas ought to be seen by everyone else in order for the community to improve. Our confession, instead, helps suggestions bubble up in a bottom-up manner. It ensures the following three things:
- Information safety. Posts are completely anonymous and hidden from viewers and even the administrators.
- A safe and friendly web space. A specific confession wall is only accessible within members of a certain community. We also included automatic toxicity detection. Toxic and other suspicious contents will be labeled and hidden from users automatically by our NLP model until an admin looks at it and marks it otherwise.
- Engaging interface. We wanted to replace that brown suggestion box with a much more appealing web application. Inspired by Pinterest, our webpage will be really enjoyable to view and fun to interact with. We hope members of a community will never get bored using our platform.
How I built it
- Web development: Django + BootStrap4
- NLP model: PyTorch
Challenges I ran into
First, toxicity detection is a challenge to us because simple rule-based algorithms didn't seem to work well in all cases. Therefore, we took a Machine Learning approach and trained a deep learning neural network model to detect toxicity automatically. We used the Kaggle Toxicity Detection link dataset and the multi-head Transformer model architecture. The resulting model reaches pretty high accuracy and recall. Another biggest challenge that we ran into was that we used Django built-in format for our form (aka the submission box). It worked perfectly fine and Django saved us lines of codes by providing the format; however, it is not as aesthetically pleasing as we want it to be. It caused us a lot of problems reading source codes and replacing the original rendering statement to finally pass the CSS variables and achieve this decent-looking website feed.
Accomplishments that I'm proud of
We are proud of finishing this project in only two days. We utilized the Django Model View Template, which is a sustainable program structure that will allow us to accommodate hundreds and hundreds of users in the future. We will put the website into use after the Bid Red Hack.
What I learned
Throughout this project, I had a better understanding of how to create a website that is able to take users’ information and render it on the website. I also learned how to use libraries like Bootstrap and Django to make building the project a lot easier and faster.
What's next for Big Red Confession Wall
We will definitely publish the website since it is working perfectly fine. We will also polish the website a little more to make it more aesthetically pleasing. Since we have no users right now to achieve the anonymousness feature, we will see if it is the best way for maintaining the website and go from there.