RoastReviews

Yelp reviews are helpful, but reading through them is a strugle. RoastReviews solves this. We made an app that uses AI to analyze reviews for coffee shops and summarizes the relevant information.

Comment

If the link is not working you can also check out our video here! https://www.loom.com/share/1ceb7d82b47744098a518fa449df7bef

Problem

Are you a coffee lover? Have you ever traveled to a new city and spent hours on google maps reading reviews to decide where to get a good cup of coffee? With so many options available, knowing which coffee shops are worth your time and money is hard. Yelp reviews can be helpful, but reading through them is time-consuming and overwhelming.

How we solve the problem

That's where our app comes in. We used AI to read Yelp reviews and aggregate them to get a rating for each coffee shop in a few relevant categories, such as food, staff, and atmosphere. Our app makes it easy to quickly find the best coffee shop in town based on your preferences.

System design

We leveraged Cohere's API to analyze Yelp reviews by feeding it into a generative language model. The model read each review and extracted sentiment categories. Our system aggregates the model’s results into a 1-5 rating for each category. In our hyperparameter tuning, we updated the temperature level, which reduced the randomness of the responses. Results are saved as a JSON file, segmented by business, and displayed via Streamlit.

Challenges + Future Work

We faced some slowdown due to the API call quotas of cohere’s API but we’re grateful for Ivan for allocating some credits for us to run our data analysis with a larger sample size. For future work, we plan on including location information about the user and custom query categories to filter our coffee shops in a personalized way. Additionally, we would like to connect the system to the yelp api to do adaptive learning and update our ratinfs as more data is created. Finally, we also intend to further learn from our data by extracting trends and performing clustering.

Target Market

Coffee is what makes us feel at home to us but not everyone cares about the same things. Our target audience is every coffee or tea lover that is new to a new city and wants to find the perfect coffee shop experience based on specific criteria.

Conclusion

We believe that finding the perfect coffee should not take time our of your trip, it should be a seamless and personalized experience. With the power of AI we found the solution for giving users the perfect coffee shop recommendations.

Built With

Share this project:

Updates