Our homepage featuring the AccurRate macot - Greg the Goose!
Our homemade ML model predicting what the hotel's rating should be based on what you wrote.
Rating different aspects of the hotel stay and having AI use the information to write a full review.

Inspiration

When you write a review for a place, such as a hotel, you have many thoughts. However, with many conflicting thoughts, it's hard to give a number from 1 to 5 for your review. With our web app, you can just write your thoughts and a machine learning model will give you a number from 1-5 stars for your review.

Moreover - if you know how you want to rate your stay at a hotel but don't feel like putting it into words for a review - we've got you covered too! Let our AI write your review for you based on how you rate certain aspects of your stay.

How we built it

For the NLP ML Model, the data would go through the following process to train the model:

Input Data: Review text, Output Data: Rating Number (1-5)
The review text is extracted and in encoded using the embed English models provided by Cohere.
We brought the encoded results into a Support Vector Machine model to predict the rating number
We tuned hyperparameters such as C, gamma and kernel in order to increase accuracy in our model.

Challenges we ran into

One of the challenges we ran into was when we attempted to use a K means classifier to implement the NLP ML model for predicting the rating number. This method is unsupervised, which meant it would have been very difficult to track the relationships between the rating and the review text. We learned how it's important to think about design and to plan our solutions out before spending lot's of time on debugging and the implementation.

Accomplishments that we're proud of

We were proud that we learned how to use AI and Taipy within such a short timespan. It was difficult to find a project that we found interesting, but we learned a lot through going to workshops and reading documentation. We went out of our comfort zone and we are glad that we explored new tools and concepts in programming.

What we learned

We learned how to take user input text and embed the text and apply machine learning
We learned how to use markups to display information in a fully functional website with Taipy
We learned the importance of knowing what we're working with and what their purpose is because without knowing these, you may end up misusing your time and effort

What's next for AccurRate

We want to be able to detect the key words in the user inputs which influenced the final rating the most and display it. We also want to improve the visual experience for the user by making the UI better and adding transitions for better ease of access.

Shout out to Taipy!

This project was our first introduction to Taipy, and we had a lot of fun learning it! We tried to bring Taipy to its limits - using multiple interactivity features (buttons, text inputs, sliders) and also extending it by writing custom styling.