Towards a Content-Based Process Mining Approach in Personal Services

Personal Services are directed at enhancing a client’s mind or body and often require flexibility for reaching service delivery goals. While often unstructured text data is created as part of this work e.g. due to documentation duties, existing information systems and process analysis methods do not leverage this process-related knowledge asset. As a technique for automatically generating process models, Process Mining mainly considers activities to visualize and analyze processes. Especially in Personal Services, the activities by itself provide too little information for a helpful model. The available unstructured text data enable refining process models by using the hidden process knowledge. Topic Modeling techniques can be used to extract the knowledge. This paper provides an approach for generating content-based process models in Personal Services. First, the text data is structured and summarized into labels by using the Labeled Latent Dirichlet Allocation (L-LDA). Thereupon, Process Mining techniques generate a process model showing the order of handled contents represented by the labels. This approach is presented and evaluated by a case study in a social care company. Problems have been identified in terms of perceived correctness and usability, while the automatic labeling process showed good results. In general, a good utility can be assumed if the identified problems are going to be resolved.