[Team Project] Automating Information Retrieval and Knowledge Extraction using NLP for FSD Administration and Production

Development of pipelines will be established to extract relevant documents from external sources (car manufacturers), internal wiki pages and documents stored on INTERGATOR.de.

Implementation of semantic search techniques to enhance contextualization and retrieval accuracy.

Integration of open-source Large Language Models (LLMs) that support both German and English to facilitate comprehensive information understanding and knowledge extraction.

We will explore the possibility of fine-tuning the LLM based on extracted and preprocessed data to improve its domain-specific performance.

The fine-tuned model will be deployed on a local server to conduct testing and evaluation of its effectiveness.

Mechanisms will be instituted to enable the LLM to output inference prompts with reference resources, as well as perform internet searches for further information gathering, if needed.

Betreuer: Karsten Wendt