A Method Based on a New Word Embedding Approach for Process Model Matching

Introduction

Process model matching(i.e., PMM)is an activity widely conducted in organizations to align process models. This alignment is critical for many business process (i.e., BP) management tasks such as storing, merging, clustering or querying Business Process models (i.e., BPs) (La Rosa et al. 2013; Goncalo et al. 2015; Weidlich et al. 2012).

Technically an alignment is a set of correspondences between activities of two process models. Each correspondence is a pair of two semantically similar sets of activities. The first set of a pair contains activities from the first BP and the second set of this pair contains activities from the second BP. Formally, an alignment is a set of not overlapped matches pairs (i.e., correspondences) {(A₁₁, A₂₁),(A_1.2,A₂₂),...,(A_1n,A_2n)}. Each pair defines a match (i.e., correspondence) between a set of activities, A_1i, from BP₁ and a set of activities, A_2i, from BP₂. Two pairs (A_1i, A₂i) and (A_1j, A_2j) does not overlap iffA_1iA_1j= and A_2iA_2j=. A correspondence (A_1i, A_2i) between a set of activities A_1i from one BP and a set of activities A_2i means that the activities A_1i and A_2i refer to the same activity in the domain. Figure 1 presents of an example of an alignment between two process models.

Figure 1.

An example of an alignment

Correspondences between the activities are presented using the grey shades. Examples of correspondence is between the set {prepare answer, applicant registration, send notification} in the process A and the set {reject students, accept student} in the business process B.

The objective of any process model matching technique is to find automatically such correspondences.

Many approaches have been proposed to achieve this objective(Goncalo et al. 2015; Daniel et al. 2017, Xue, 2019; Khurram et al. 2019). The proposed approaches are widely based on a combination of lexical, syntactic and semantic similarity metrics proposed in different fields such as the NLP (i.e., Natural language Processing) community (Goncalo et al. 2015; Daniel et al. 2017).