NLP-assisted troubleshooting
Project team:
- John Pavlopoulos, Affiliated Researcher, DSV, SU
- Tony Lindgren, Associate Professor, DSV, SU
- Alex Xenos, Research Assistant, DSV, SU
Project period: 2022-02-01 to 2022-07-31
Funding source: DSV internal project
Budget: 395K SEK
Description
Fault diagnosis is the task of detecting the fault that caused a problem or unexpected behaviour to a subject. By exploring Artificial Intelligence, and especially Natural Language Processing (NLP) solutions, this project will (i) assist trustworthy decision-making, (ii) support professionals, and (iii) optimise diagnostics tasks. The focus will be on troubleshooting management for companies that own hundreds or thousands of mechanical systems (e.g., robots or heavy duty truck fleets). Any fault claim (usually written in natural language, such as emails, SMSs, etc.) arriving at the company’s front desk will first be passed through a supervised neural text-classifier (the SoTA in several NLP tasks), trained to detect the true fault behind this claim. At this stage, neither the end user (e.g., the driver) nor the company knows the problem. The assumption, however, is that an AI system can learn to predict the underlying fault by only reading the textual claim. The system-predicted fault, then, can assist the mechanics/diagnostics teams to reach faster to the root cause of the problem and achieve speedier troubleshooting.
People
John Pavlopoulos
