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

This paper proposes a method based on a new word embedding approach for matching business process model. The proposed method aligns two process models in four steps. First activity labels are extracted and pre-processed to remove meaningless words, then each word composing an activity label and using a semantic similarity metric based on WordNet is represented with an n-dimensional vector in the space of the vocabulary of the two labels to be compared. Based on these representations, a vector representation of each activity label is computed by averaging the vectors representing words found in the activity label. Finally, the two activity labels are reported as similar if their similarity score computed using the cosine metric is greater than some predefined threshold. An experiment was conducted on well-known dataset to assess the performance of the proposed method. The results showed that the proposed method shared the first place with RMM/NHCM and OPBOT tools and can be effective in matching process models.