Diagnosing Data Pipeline Failures Using Action Languages

This paper discusses diagnosing industrial data processing pipelines using action languages. Solving the problem requires mechanisms for reasoning about actions, their effects and mechanisms for accessing outside data source. To satisfy these requirements, we introduce an action language, Hybrid ALE that combines elements of the action language Hybrid AL and the action language CTAID. We discuss some of the practical aspects of implementing Hybrid ALE and describe an example of its use.

• Algorithms and Theory

• Machine Intelligence