NIT: A dataset for network intent translation

Area of Science:

• Computer Science
• Network Engineering

Background:

• Network configuration is complex and often manual.
• Automating network management requires translating high-level intents into low-level configurations.

Purpose of the Study:

• Introduce the Network Intent Translations (NIT) dataset.
• Facilitate research in network intent translation, specifically intent-to-vendor translations.

Main Methods:

• The NIT dataset comprises 1000 entries, each a JSON object with question (intent), context, and answer (configuration).
• Data was generated and validated for Juniper EX3300 switches running JUNOS, ensuring syntactic correctness.

Main Results:

• The dataset provides natural language intents mapped to Juniper device configurations.
• It focuses on translation accuracy, not parameter value validation or conflict resolution.

Conclusions:

• The NIT dataset is a valuable resource for advancing network intent translation research.
• It supports the development of automated network management solutions.