Conversational agents have the potential to streamline tasks, provide support, and enhance user experience across various domains including Virtual Research Environments (VREs). The recent progress in conversational artificial intelligence and large language models (LLMs) offers novel strategies for the development of these agents. Janet is an attempt to develop an agent that, by leveraging the resources within the VRE, can engage in natural language conversations with VRE users to help them manage their daily activities, find relevant information, and use what the specific environment offers. It is developed following the Retrieval-Augmented Generation paradigm, a technique that reduces the effect of one of the limitations affecting LLMs; namely, hallucination. This talk highlights the lessons learned during the development of Janet.
Introducing Janet: early findings on a conversational agent for Virtual Research Environments
Candela L.
2024
Abstract
Conversational agents have the potential to streamline tasks, provide support, and enhance user experience across various domains including Virtual Research Environments (VREs). The recent progress in conversational artificial intelligence and large language models (LLMs) offers novel strategies for the development of these agents. Janet is an attempt to develop an agent that, by leveraging the resources within the VRE, can engage in natural language conversations with VRE users to help them manage their daily activities, find relevant information, and use what the specific environment offers. It is developed following the Retrieval-Augmented Generation paradigm, a technique that reduces the effect of one of the limitations affecting LLMs; namely, hallucination. This talk highlights the lessons learned during the development of Janet.File | Dimensione | Formato | |
---|---|---|---|
IWSG 2024 Slides.pdf
accesso aperto
Descrizione: Slides
Tipologia:
Altro materiale allegato
Licenza:
Creative commons
Dimensione
1.11 MB
Formato
Adobe PDF
|
1.11 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.