Learn Lugano: generative AI for efficient information retrieval
Client: Lugano Living Lab, Città di Lugano
Industry: Education
7
Portals involved
+40k
Web pages extracted for bot
1
Website for all needs
Learn Lugano is an online platform created through a collaboration between the city of Lugano and the city's major universities, designed to simplify access to information for current and prospective students and to promote Lugano as a university destination. Given the amount of information available, it was necessary to develop and integrate Lucy, an advanced AI-based generative chatbot designed by Ander Group. By scraping the sites of USI, SUPSI and Franklin University, Lucy provides accurate and up-to-date answers, enhancing the user experience and facilitating information searches.
GPT-4
Langchain Framework
Retrieval Augmented Generation
How to navigate through more than forty thousand web pages?
In a context such as the university, where information relevant to students is often fragmented across several portals, it can become complex and time-consuming to navigate the various resources for quick and accurate answers.
The amount of data available, from course information to administrative details and student life opportunities in Lugano, requires a centralized and easily accessible solution.
Learn Lugano was created precisely to meet this need, serving as a collector for all this information.
However, even on a site where all the information is centralized, a solution was needed to make the search even more efficient and immediate.
Hi there, I’m Lucy.
How can I help you?
Out of this need came Lucy, an advanced chatbot based on generative AI.
Indeed, during the design phase, the need matured to provide users with a conversational interface that would allow them to search and access available information more efficiently.
Lucy not only gives direct responses by engaging users in natural conversation, but she also offers direct links to relevant pages on Learn Lugano and other portals, such as USI, Ticino Turismo, SUPSI, Franklin, and Lugano.ch.
This is done on the basis of the pertinence of the question asked to the content of the pages, thus improving the user experience and making it easier to access relevant information.
The chatbot we created is based on the OpenAI GPT-4 model, with API integration, and uses the retrieval-augmented generation (RAG) mechanism. The very first step of this method involves retrieving relevant content. This content is then fed at runtime into the generative model, along with user's prompt. This approach helps the system answer users’ questions more accurately and pertinently.
It was essential to perform a preliminary web scraping process in this phase, which consisted in extracting content from the HTML pages of the various portals for inclusion in the chatbot.
During the development process, we also drafted guidelines for Lucy’s answers using so-called “system prompts”. This document sets out the assistant’s role and the situations in which she can help.
The final product is therefore an integral tool within the website that can offer an alternative to conventional website navigation by means of a fluent and natural conversational interface. This allowed us to eliminate the need to navigate between various predefined pages and categories to find the desired information, streamlining the user experience enormously as a result.
"Learn Lugano is a project we strongly believed in and have been involved with from the very beginning. Through the integration of useful information for students and the adoption of innovative technologies like the AI-powered chatbot Lucy, we are shaping a future where academic experience and the natural beauty of Lugano harmoniously blend, offering a unique and inspiring environment for study and personal growth."
Robert Bregy / City Secretary, Città di Lugano