Ashmi Banerjee
updates
32cae45
raw
history blame contribute delete
3.5 kB
<!-- intro.html -->
<body>
<h1 style='font-size:xx-large; color: green; text-align: center'>🍀 Green City Finder 🍀</h1>
<h3 style="text-align: center">AI Sprint 2024 submissions by Ashmi Banerjee.<sup>*</sup></h3>
<br>
<p style="text-align: justify">
Tourism Recommender Systems (TRS) have traditionally focused on providing personalized travel suggestions, often
prioritizing user preferences without considering broader sustainability goals.
Integrating sustainability into TRS has become essential with the increasing need to balance environmental impact,
local community interests, and visitor satisfaction.
We enhance the traditional RAG system by incorporating a sustainability metric based on a city’s popularity and
seasonal demand during the prompt augmentation phase.
This modification, called <b>Sustainability Augmented Reranking (SAR)</b>, ensures the system's recommendations align with
sustainability goals.
</p>
<p style="text-align: justify"><a href="https://arxiv.org/pdf/2403.18604">Sustainability score</a> for the retrieved
destinations is calculated based on the following parameters:
<ul>
<li>Carbon footprint from the starting points to the retrieved cities using the greenest mode of travel (fly, drive,
train)
</li>
<li>Overall popularity of the retrieved destinations based on their aggregated Tripadvisor reviews and opinions</li>
<li>Seasonal footfall for the intended month of travel (if present)</li>
</ul>
</p>
<p style="text-align: justify">
We test our implementation with Google's <b>Gemini</b> models
through VertexAI to generate sustainable travel recommendations.
We use the Wikivoyage dataset to provide city recommendations based on user queries.
The vector embeddings are stored and accessed in a VectorDB (LanceDB) hosted in Google Cloud.
</p>
<p style="text-align: justify">This is an extension of the following work. To <b>cite</b>, please use the following:</p>
<blockquote>
<p> [1] <b>Enhancing sustainability in Tourism Recommender Systems,</b> <i>Ashmi Banerjee, Adithi Satish, Wolfgang
Wörndl</i>, In Proceedings of the 1st International Workshop on Recommender Systems for Sustainability and Social
Good (RecSoGood 2024), co-located with ACM RecSys 2024, Bari, Italy.
</p>
</blockquote>
<blockquote>
<p> [2] <b>Modeling Sustainable City Trips: Integrating CO2e Emissions, Popularity, and Seasonality into Tourism Recommender Systems,</b> <i>Ashmi Banerjee, Tunar Mahmudov, Emil Adler, Fitri Nur Aisyah, Wolfgang
Wörndl</i>, arXiv preprint <a href="https://arxiv.org/abs/2403.18604">arXiv:2403.18604 (2024)</a>.
</p>
</blockquote>
<br>
<p style="text-align: justify; font-weight: bold"><sup>*</sup>Google Cloud credits are provided for this project.</p>
<h2 style='font-size:large; color: black; text-align: left'>Instructions</h2>
<ul>
<li>Select the country and city where you're located.</li>
<li>Enter the search query; it has to be something for which the system can recommend cities.</li>
<li>Click the <b>Search</b> button to find the most sustainable recommendations for your <b>starting
position</b>.
</li>
<li>Click the <b>Clear</b> button to clear the fields.</li>
</ul>
<p style="text-align: justify; color: darkred">Note that this works best if you ask it for <span
style="font-weight: bold; color: darkred; text-underline: darkred">city</span> recommendations.</p>
</body>