General feedback

[mari-s4e]

Several s4e collleagues tried the chatbot v1. To summarise the feedback:

• answers are too long
• basic questions about the project (e.g. project duration, funding, who to contact etc...) cannot be answered ("I'm sorry, but the provided context does not contain specific information about "
• answers are too generic, e.g. for questions like "What processing methods are used to evaluate Urban Adaptation to Climate Change?", "what is the validation routine for datasets ingested in fairicube?", "How is Fairicube related to the EU Habitats Directive?", "how to login to the fairicube platform?"

Finally, looks like we are developing a mobile banking application now!

[gmartirano]

Dear @mari-s4e,
thank you for your feedback and apologies for the very late answer, but, not being tagged, we left it out from our radar (our mistake).
Regarding the issue of long and generic answers, we described the know reasons in issue#1.
Regarding the reply to the question “What is Use Case 1 about” that you pasted, this is a typical well know case of “hallucination” which can occur when the model generates answers based only on its pre-trained knowledge. We mitigated this risk using GPT-4o LLM (large language model), as the generative part responsible for producing coherent and contextually relevant responses. And, indeed, in order to properly leverage such ChatBot feature, it would be enough to better formulate the question, e.g. “What is the FAIRiCUBE Use Case 1 about?”. In this case the (long) answer is very coherent and contextually relevant 😊.
Due do the ChatBot experimental characteristics, we will add some tips in the UI, to better guide the ChatBot users and (try to) minimise their disappointment 😊.