Enabling natural language analysis for object-centric event logs

Abstract

Object-centric process mining is an emerging family of techniques for analyzing business processes where events involve multiple interconnected objects. However, formulating queries for object-centric event logs might be challenging when using traditional query languages such as SQL, as these languages typically demand the user to have programming expertise for handling the intricate relationships inherent in object-centric data. To address this challenge, we introduce a conversational framework designed to facilitate the analysis of object-centric event logs following the OCEL 2.0 format. By leveraging the capabilities of Large Language Models (LLMs) with Retrieval Augmented Generation (RAG), our framework enables users to interact with object-centric logs in natural language, eliminating the need for conventional query languages. Additionally, we present a dataset derived from a SAP Procure-to-Pay event log as a benchmark for assessing the performance of conversational techniques in analyzing object-centric event logs from various perspectives. The evaluation demonstrates that our framework enhances the accessibility of OCEL 2.0 log analysis, providing accurate and faithful responses.