Activity Finder is a web application built with Next.js and Tailwind CSS. The app allows users to search for activities in a specific city by utilizing a custom-built API and a simple crawling mechanism. It combines the power of Natural Language Processing (NLP) and geospatial data to deliver relevant activity search results.

This project aims to help users easily discover activities, events, and things to do in various cities. The core functionality revolves around an API that processes a set of JSON data files, extracts useful information, and uses the TF-IDF technique to deliver accurate and relevant results based on user queries.

The app also integrates Leaflet, a leading JavaScript library for maps, and OpenStreetMap for geocoding.

• Frontend: Next.js, Tailwind CSS
• Backend: Next.js
• Libraries/Tools: Leaflet, OpenStreetMap, TF-IDF Algorithm
• Data Storage: Local JSON files (scanned from the src/db/ directory)

The core of the app resides in the /api/lookup endpoint, which handles the user queries and retrieves activity information using the following process:

1. Crawling:

   • The first query triggers the server to perform a crawl. It scans the src/db/ folder for all .json files that are formatted as listX.json.
   • The server reads each JSON file, concatenates all the activity data, removes duplicates, and processes the content.
   • The server then uses the TF-IDF technique to calculate and store word statistics in a file called tfidf.json.

2. TF-IDF Storage:

   • After the first request, the tfidf.json file is saved to disk, and the server will use the data from this file for subsequent requests.
   • If the server restarts and the tfidf.json file already exists, the server automatically loads the data from the file without re-crawling the .json files.

   This behavior is not highly scalable but is suitable for the current project size. The intention is to avoid setting up a full database and instead leverage local JSON files for simplicity.

3. Query Handling:

   • For each query, the server uses the TF-IDF technique to search for the most relevant activities. It compares the query terms with the stored statistics to find the best results.
   • The filtered and ranked results are sent back to the frontend for display.

Once the results are returned from the server, the frontend displays them in an easy-to-read format. Additionally, Leaflet is used to show maps, and OpenStreetMap is used to convert addresses into geographic coordinates for accurate location mapping.

• Search for activities by city or location.
• Display results on a map using Leaflet and OpenStreetMap.
• Uses the TF-IDF technique to filter and rank results based on relevance.
• Caches and reuses the TF-IDF data for faster subsequent queries.

You may find a demo of the website deployed on Vercel here.

Some example inputs to try: surf plage parc activités de sport

Feel free to fork the repository and submit pull requests!

This project is licensed under the MIT License - see the LICENSE file for details.

If you have any questions or feedback, feel free to open an issue or contact me!