Business Processes Retrieval based on Multimodal Search and Lingo Clustering Algorithm

Abstract

This paper presents a model for searching and grouping of business process models. To search business process models, the model contains a module for multimodal indexing that takes into account textual and structural information of models. To group models, an adaptation of the Lingo algorithm was used; it is based on singular value decomposition and frequent phrases extracted from the description (textual and structural information) of the business process models retrieved according to a query. The evaluation of model was conducted by executing the search process on a closed test-collection built collaboratively, that containing 146 business process models, and comparing the results with a set of relevant business process models obtained from an evaluation issued by 59 experts. Measures of relevance such as graded precision and recall shown promising results in the search process, as well as the quality assessment of the sets formed by the grouping process.