From free-text repair action messages to automated case generation

Free-text problem description messages and repair action messages, which are written down in airplanes log books by pilots and technicians, contain very valuable information about problem symptoms, the pieces of equipment involved in the repair and the actions performed on them, and the results of those actions on the symptoms. This information could be used by a diagnostic system to provide suggestions about the nature of the problem and the proper repair actions. We use natural language analysis and interpretation techniques to extract that information from the messages with an aim at using it to generate automatically new cases for the case-base reasoning performed by the IDS system at Air Canada.