Llm-Basiertes System und Verfahren zur Profilierung von Chatbot-Benutzern

Details

Veröffentlichungs-Nr.

EP4769206

Anmeldetag

29. Dezember 2025

Veröffentlichung

1. Juli 2026

Rechtsraum

EP

IPC

G06F40/30, G06N3/08, G06N5/022

Abstract

An LLM-based method for profiling chatbot users' technical proficiency and performing adaptation in the chatbot response, comprising the step of receiving encountered problem by the troubleshooting chatbot via a user prompt; using a profile inference LLM to obtain inferred profiles of chatbot users' technical proficiency regarding cybersecurity troubleshooting, by analyzing current prompt as entered by the user, from a data storage; the most recently updated user profile, which encapsulates the past user data; troubleshooting-oriented cybersecurity taxonomy consisting of proficiency domains and subdomains; a collection of inferred profiles of all users; hardware, software or networking-related data collected from the user's device; generating a vector of profile scores that quantify troubleshooting-oriented cybersecurity proficiency subdomains; updating the user's inferred profile by sending the inferred profile to the collection; replacing the older profile of the user by an updated profile; sending the inferred profile to a response adaptation LLM, along with the response generated by the chatbot; adapting the technical instructions, produced by the chatbot to the inferred user's profile; sending the profile-adapted response to the user, to follow the chatbot's instructions.

Anmelder

Firma

Deutsche Telekom AG

Land

🇩🇪 Deutschland

Fachgebiete