LLM-Based Network Label Transformation for End-to-End Visibility
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Communication networks with multi-vendor and multi-generational devices face challenges in obtaining end-to-end service path visibility due to inconsistent vendor-specific network management systems and unstructured human input labels, which are difficult to decipher and often stored inconsistently, leading to incomplete inventory databases and missing neighbor connections.
Innovation Solution
A system that uses a large language model (LLM) to interpret and transform unstructured human input labels, combining them with inventory data to determine neighbor connections and create a standardized data model, enabling end-to-end service visibility by stitching together service ports in a multi-vendor/multi-generational network management system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If vendor-specific network management systems are used for each network element, then device management capability is provided, but end-to-end service path visibility is lost due to inability to communicate across different vendors and generations
Solution Approach 1:
A network management system acts as an intermediary that collects unstructured human input labels from multiple vendor-specific network elements and uses an LLM to interpret and transform them into a unified structured format. This mediator enables end-to-end service path visibility across multi-vendor networks without requiring each device to communicate directly with others.
Solution Approach 2:
The system changes the parameter format of network element identifiers from unstructured human input (varying formats across vendors) to a standardized structured format. The LLM transforms different vendor-specific label formats into a common representation, enabling consistent service path visibility across heterogeneous networks.
2Ease of operation
If unstructured human input labels are used for network element identification, then flexibility in labeling is provided, but data processing and interpretation difficulty increases
Solution Approach 1:
The manual mechanical process of parsing and interpreting various unstructured label formats is replaced by an automated LLM-based system. The LLM processes diverse human input labels and transforms them into standardized structured data, eliminating the complexity of manual interpretation while preserving the flexibility of human-labeled data.
Solution Approach 2:
The system transforms the parameter representation from unstructured human-readable labels to structured machine-processable format. The LLM converts various label formats (different vendors, generations, and styles) into a consistent structured representation, making data processing straightforward while maintaining input flexibility.
3Loss of information
If human labels are manually entered for each network element, then service path identification is possible, but inventory database completeness is reduced due to irregular and inconsistent labeling
Solution Approach 1:
The manual process of consistently labeling and indexing network elements is replaced by an automated LLM system. The LLM processes unstructured human input and generates consistent structured identifiers, dramatically improving inventory database completion efficiency while preserving complete service path information.
Solution Approach 2:
The system transforms inconsistent manual labeling parameters into standardized structured parameters. The LLM normalizes various labeling conventions into a unified format, ensuring complete and consistent inventory database entries without requiring manual standardization effort.
4Adaptability or versatility
If legacy network elements are supported in multi-generational networks, then network compatibility is maintained, but service path visibility is lost due to lack of manufacturer support
Solution Approach 1:
The LLM-based network management system serves as an intermediary that bridges legacy network elements with modern management systems. It collects unstructured labels from legacy devices that lack official manufacturer support and transforms them into standardized formats, enabling reliable service path visibility for obsolete equipment.
Solution Approach 2:
The system provides universal support for multi-generational network elements through a single LLM-based platform. The same system handles both legacy and modern devices, translating their different labeling conventions into a unified representation, thereby achieving both compatibility and visibility.
Data Source
AI summary
Some embodiments enable the interpretation and transformation of unstructured human input to enable end-to-end service visibility across a transport network. Some embodiments include assembling and sending a prompt to a large language model (LLM) to interpret and transform examples of unstructured human input describing a circuit to network element (NE) information, where the NE information can be used to identify neighbor devices between different vendor products providing a service. The LLM can output data points that enable identifying neighboring devices on an end-to-end service path that includes multi-vendor and/or multigenerational devices. The data points used in conjunction with inventory data enables neighboring connections to be determined so that islands of different multi-vendor and/or multigenerational devices can be connected to neighbor devices along the end-to-end service path. The neighboring connections can be added to an inventory database enabling visibility end-to-end service path.