Inspiration
People love to travel a lot to new countries and cities. However, finding places you would like can be challenging because everything is so new. So, we developed Tripseeker to help recommend places and create an itinerary based on your past Google Maps data.
What it does
Tripseeker takes in either your Google Maps Takeout or profile. Based on your travel location or your desired area of interest, we generate a full travel itinerary tuned to your personal interests and past experiences. Further, we revolutionize sightseeing once and for all. Combining good food, good sights, and good activities based on YOUR preferences.
How we built it
We built the backend, grabbing data from users' Google reviews. Then, based on the desired location and search settings, we send it to our deep learning pipeline. Based on this, we tokenized ratings using DistellBERT. With the embeddings, we pass them in our RAG pipeline. Finally, with our highest compatibility locations, we pass them into our fine-tuned LLM for itinerary planning.
Challenges we ran into
A challenge we ran into was collecting data from Google Maps, which we overcame by using a mix of Selenium, JavaScript, and BeautifulSoup. We also ran into difficulty making the recommendation model fast, so we used multithreading to improve performance. We also ran into merge conflicts nearly every time we pushed, which we had to resolve. We also found it difficult to design the homepage and settled on one that worked.
Accomplishments that we're proud of
We are proud of developing an algorithm that can tell what a person likes and does not like purely based on the reviews they gave on Google Maps, and we could then use that data to give further recommendations in new places. We are also proud of designing a clean, appealing UI for the website.
What we learned
We learned how to analyze a person's preferences and dislikes based on the reviews they give of places they visit. We also learned how to make a clean website UI and work together as a team, especially resolving merge conflicts.
What's next for Tripseeker
We hope to further develop our model to make it more consistent and accurate for better recommendations and ultimately a better traveling experience. Tripseeker can also be expanded to mobile apps for easier accessibility.
Log in or sign up for Devpost to join the conversation.