Fault diagnosis is crucial in identifying the root cause of problems, whether in medicine or automotive. Our study at Stockholm University, led by Tony Lindgren, John Pavlopoulos, and Korbinian Randl, explored using Natural Language Processing (NLP) for multilingual automotive fault diagnosis.

Using data from Scania, we analyzed textual descriptions from work orders to classify vehicle malfunctions. We fine-tuned a multilingual Transformer model, achieving better results than current industry standards.

This method isn’t just for automotive applications; it can be used in any industry where fault claims are textual, like agriculture.