www.thedomainisalie.tech

Employ the power of magic to make predictions about whether your reviews are positive, negative or neutral!

Comment

Inspiration

We are all new to the world of Hackathons - and to building things generally - thus our inspiration for the weekend was simply to learn new stuff and try to build something cool. All of us liked the theme and aimed to make use of AI in our project.

What it does

This website is similar to how someone would use a magic 8 ball. The user asks a yes or no question then clicks on the button to receive their fortune!

How we built it

In the beginning, we tried using Python and JavaScript to access Cohere, which uses some of our training data to classify product review messages as either positive, negative, or neutral. We didn't manage to integrate these with our website though, so we switched to our fortune telling idea instead. (You can see our Cohere attempts by switching to different branches in the GitHub repo). We ended up hosting a static website on AWS with a domain name from domains.com.

Challenges we ran into

  1. Unclear ideas - our aim was 'make something AI', which was a fairly broad topic and we would have benefitted from getting a clearer direction, sooner.
  2. Low experience - none of us had done Hackathons before, only one of us had anything like AI, so some more background knowledge would have been helpful.
  3. Time zones! - minor one, but we all worked in different time zones (PT, ET, GMT) so had to be flexible with time.
  4. Google Cloud - we couldn't find a way to connect the Domain.com registration to Google Cloud authentication.
  5. Cohere code - we ran into significant errors with the code that used the Cohere API, and could not make this work properly in either Python or JavaScript.

Accomplishments that we're proud of

Josh: proud of getting the website hosted in AWS and configuring the domain so the shortened version pointed to the bucket correctly. Vivian: proud of overcoming all the challenges faced during the learning process and building most of the front-end

What we learned

How to cope when it all goes a bit wrong! As a team, we developed more understanding of Hackathons and what to aim for, plus resilience and creativity from pivoting and changing direction. Josh learned a lot from our false starts: 1. Java basics and getting things setup, plus how to run code with 'java' vs 'javac'; 2. API basics and how to make connections; 3. Cloud website hosting, both practically for a static site and theoretically for dynamic sites. Vivian also learned a lot from false starts: 1) k-means clustering 2) intro to neural networks (3blue1brown <3) 3) Figma 4) Cohere basics 5) the cool things html & css can create; this was my first time making things like hover animations and SVG waves.

What's next for our project

With more time, we would have creates a Flask server to run the back-end Python script, and host the website such that it supported dynamic updates. We would also have made the Cohere model better, or possibly tried to implement our own simplified version of a learning model instead. Ideally we could also find a way to push our GitHub updates to AWS automatically.

Built With

Share this project:

Updates