Unlocking Telecom Domain Knowledge Using LLMs

Abstract

Conversational assistants have become increasingly popular as they use Large Language Models (LLMs) and Retrieval Augmented Generation (RAG) for domain context. In this work, we present an end-to-end solution that leverages RAG for telecom domain Question Answering (QA) on standards documents. We highlight that retrieval quality is important, along with an efficient indexing mechanism for the document embeddings. We also index images and tables for QA on standards documents. Our Telecom Knowledge Assistant is useful for handling specific queries from telecom domain experts, as well as for novice learners. The developed approach and solution are amenable to adapt for other domains as well.