Adapting Foundation Models for Operator Data Analytics
Manikanta Kotaru
Abstract
The complexity of operator networks and myriad of specialized metrics produced by network function providers present a formidable challenge in retrieving and analyzing operator data, a vital component for network operations. This necessitates specialist intervention, which is time-consuming and limits customization. This paper proposes Data Intelligence for Operators Copilot, a natural language interface for retrieval and analytics tasks on operator data, leveraging foundation models. It addresses the challenges posed by operator data through a novel application of semantic search to effectively provide necessary context regarding specialized metrics. The system has outperformed state-of-the-art natural language interfaces for databases, when applied to an operator-specific benchmark dataset of expert-generated representative queries, with 66% execution accuracy.