SentiNet: Semantic-Driven Intent Networking for SDN

Applying Large Language Models (LLMs) to network configuration has the potential to reduce the need for specialized expertise and simplify the configuration process. This paper introduces SentiNet, a network framework that integrates an LLM into Software-Defined Networking (SDN) to realize intent-based networking by translating natural language network intents into low-level network configurations. To enhance the accuracy of intent decomposition and configuration generation, we employ a multi-turn dialogue approach and a keyword-based example indexing mechanism. Experimental results validate that the proposed methods effectively enable the completion of network configurations.