Natural language-based detection of semantic execution anomalies in event logs

Highlights：

• We propose to detect anomalies in business processes based on the meaning of events.

• Our approach recognizes behavior as anomalous when it violates common-sense patterns.

• We employ state-of-the-art NLP techniques and a process-independent knowledge base.

• Our evaluation experiments reveal that our approach achieves accurate results.

• We show its complementary nature to existing, frequency-based detection approaches.