1 Mining Business Policy Texts for Discovering Process Models: A Framework and Some Initial Results Jiexun Li Drexel University jiexun.li@ischool.drexel.edu Harry J. Wang University of Delaware hjwang@lerner.udel.edu Zhu Zhang University of Arizona zhuzhang@u.arizona.edu J. Leon Zhao University of Arizona jlzhao@u.arizona.edu Abstract Many organizations use digitized policy manuals to help govern their business operations. However, many business processes are often not synchronized in a timely manner with the business policies because of the high costs of redesigning business process models in the face of frequent policy changes. As such, there is a great need for more efficient ways to maintaining the alignment between business policies and processes. In this paper, we propose a novel approach that will help process designers extract process models out of policy manuals using information retrieval techniques; we refer to our approach as Policy Based Process Mining (PBPM). PBPM can help organizations better understand their business processes by analyzing and visualizing their existing policies. In this paper, we present the PBPM framework and validate the first step of the framework by identifying process-related business policies automatically. Our initial experiments using bag-of-words and tree kernel techniques produced very promising results. Keywords: Process Mining, Business Policy, Business Process Management, Text Mining 1. Introduction Business policies enable the efficient management of an organization by defining the standard procedures and rules for its daily business operations, e.g. eleven organization-wide policies are identified in (Peltier 2004), such as Employee Standards of Conduct, Workplace Security, Information Security, Business Continuity Planning, to name a few. Among various types of business policies, many of them are used to define or constraint some aspects of the business processes, such as order fulfillment, product development, travel reimbursement, and cash handling, which we refer to as process policies. For instance, a travel reimbursement policy may define that a reasonable exception request form (RERF) must be submitted if a travel reimbursement form (TRF) is submitted later than 60 days upon completing the travel. This policy specifies the condition under which RERF submission task must be executed. In order to describe a business process, a great amount of process policies are often needed to be established. For example, we studied a major US public university’s business policy manual in our previous research (Wang et al. 2006), which has 19 sections with average 10 subsections in each section. In particular, the subsection on travel regulation by itself includes 14 topics with more than seven thousand words. Recently, in order to achieve various regulatory compliances such as Sarbanes-Oxley, organizations are investing a great amount to revamp their business policies. The process owners, executives, internal and external auditors often have to study business policy manuals that have hundreds or thousands of pages to understand the whole business processes, which is a time-consuming and overwhelming task. In addition, it is also a challenge for employees to find the information about processes to do their job correctly and efficiently due to the huge amount of policies. We refer to this problem as Process Policy