Description
In a rapidly changing environment, organizations must adapt their business processes continuously. While numerous methods enable enterprises to conceptualize and analyze their organizational structure, the task of business process modeling remains complex and time-consuming. However, by reusing and adapting existing process models, enterprises can reduce the task’s complexity while improving the quality of results. To facilitate the identification of adaptable processes, several techniques of business process similarity (BPS) have been proposed in recent years. Although most approaches produce sound results in controlled evaluations, this paper argues that their applicability is limited when analyzing real-world processes, which do not fully comply with notational labeling specifications. Consequently, we aim to enhance existing BPS techniques by using corpus statistics to account for the explanatory power of words within labels of process models. Results from our evaluation suggest that corpus statistics can improve BPS computations and can positively influence the quality of practical implications.
Recommended Citation
Fischer, Marcus; Imgrund, Florian; Janiesch, Christian; and Winkelmann, Axel, "Corpus Statistics for Measuring Business Process Similarity" (2017). AMCIS 2017 Proceedings. 1.
https://aisel.aisnet.org/amcis2017/SystemsAnalysis/Presentations/1
Corpus Statistics for Measuring Business Process Similarity
In a rapidly changing environment, organizations must adapt their business processes continuously. While numerous methods enable enterprises to conceptualize and analyze their organizational structure, the task of business process modeling remains complex and time-consuming. However, by reusing and adapting existing process models, enterprises can reduce the task’s complexity while improving the quality of results. To facilitate the identification of adaptable processes, several techniques of business process similarity (BPS) have been proposed in recent years. Although most approaches produce sound results in controlled evaluations, this paper argues that their applicability is limited when analyzing real-world processes, which do not fully comply with notational labeling specifications. Consequently, we aim to enhance existing BPS techniques by using corpus statistics to account for the explanatory power of words within labels of process models. Results from our evaluation suggest that corpus statistics can improve BPS computations and can positively influence the quality of practical implications.