Increasing Recall of Process Model Matching by Improved Activity Label Matching

Comparing process models and matching similar activities has recently emerged as a research area of business process management. However, the problem is fundamentally hard when considering realistic scenarios: e.g., there is a huge variety of terms and various options for the grammatical structure of activity labels exist. While prior research has established important conceptual foundations, recall values have been fairly low (around 0.26) – arguably too low to be useful in practice. In this paper, we present techniques for activity label matching which improve current results (recall of 0.44, without sacrificing precision). Furthermore, we identify categories of matching challenges to guide future research.