Postal address
KreativInstitut.OWL
Emilienstraße 45
32756 Detmold
Directions
Kreativ Campus Detmold
Bielefelder Straße 66a
32756 Detmold
Contact
Have you ever tried to google a problem when you don’t even know how to describe it? Or tried to find a manual for a software function when you have no idea what the function is called? In cooperation with Hirschbiegel+Grundstein GmbH, we are developing HaBiBi, the HandBookBot, and researching whether and how chatbots can help us overcome this speechlessness.
Project participants
Anna Viktoria Katrin Plaksin
Conrad Dreyer
Hizkiel Mitiku Alemayehu
Andreas Prohl-Plaksin
Project status
Testing phase
Andreas Prohl-Plaksin (Hirschbiegel+Grundstein GmbH)
While the potential is theoretically great, many questions must be resolved before it can be practically applied:
How do you integrate a chatbot into an existing infrastructure?
What legal and financial frameworks need to be considered?
How should a knowledge base be structured to ensure the chatbot provides good answers?
How can incorrect answers be avoided?
Together with the software company Hirschbiegel+Grundstein GmbH, we are developing HaBiBi and testing its use within their KOKOS.Event package.
We use the technique of Retrieval Augmented Generation (RAG).
This is a two-step process. First, a query is sent to a retrieval model to fetch relevant content from a database. In the second step, these relevant pieces of information, along with the query, are sent to a generative model. The generative model uses the retrieved documents as a knowledge source to answer the question.
This approach combines the strengths of information retrieval systems with generative models. The model searches for relevant documents or information from a database and uses them as context to generate more accurate and informative responses. This not only improves the precision and relevance of generated texts, especially in specialized or knowledge-intensive fields, but also enables a fast and resource-efficient implementation, as pre-built components are combined, eliminating the need for extensive training.