Natural Language Processing Applied to Dynamic Workflow Generation for Network Management

Workflow implementation in network management is a challenge, especially when it depends on human interaction. Traditional workflow generation, which was primarily focused on business management, employed a graphical representation of managed elements, supported using a human-understandable explanation. The difficulty with this approach is workflow adaptation to avoid execution problems when unexpected errors occur. In such circumstances, human intervention is necessary to resolve the problem. Current workflow generation solutions are inflexible and inefficient with regards to self-checking and self-healing, and they cannot be easily scaled or adapted. This paper presents a novel workflow management framework that controls an environment using dynamic workflow generation. The framework detects inconsistencies in workflow execution and reacts automatically to prevent a workflow termination or to replace a corrupted workflow if execution is not possible. A language-based text-oriented ticket input is used to describe the problem to be solved. This text is translated using natural language processing to a syntax understandable by the framework in order to generate the required workflow. The functionality of the workflow management framework is demonstrated in a software-defined networking environment.