Mining E-Mail Messages

Abstract

Increasingly information systems log historic information in a systematic way. Workflow management systems, but also ERP, CRM, SCM, and B2B systems often provide a so-called "event log" (i.e., a log recording the execution of activities). Thus far, process mining has been mainly focusing on structured event logs resulting in powerful analysis techniques and tools for discovering process, control, data, organizational, and social structures from event logs. Unfortunately, many work processes are not supported by systems providing structured logs. Instead, very basic tools such as text editors, spreadsheets, and e-mail are used. This article explores the application of process mining to e-mail (i.e., unstructured or semi-structured e-mail messages are converted into event logs suitable for application of process mining tools). This article presents the tool EMailAnalyzer, embedded in the ProM process mining framework, which analyzes and transforms e-mail messages to a format that allows for analysis using our process mining techniques. The main innovative aspect of this work is that, unlike most other work in this area, our analysis is not restricted to social network analysis. Based on e-mail logs, we can also discover interaction patterns and processes.

Multi‐perspective business process discovery from messaging systems: State‐of‐the art

“…One of these rules consists of specifying, in which kind of textual pattern, one information type can appear in an email. Van der Aalst et al 30 for example assume that the associated BP name is explicitly included in the email subject. Such an approach supposes that email interlocutors must include the BP name and the related BP elements in the email subject.…”

“…The other perspective types (data and organizational) are almost not discussed. Only Van der Aalst et al 30 have studied in EmailAnalyzer , the organizational perspective by proposing two actor involvement diagrams ; (i) task person diagram showing the relation between activities and actors, and (ii) case‐person diagram showing the relation between actors and instances. Additionally, it has studied the social network of activities' related emails' users.…”

“…A1: The presence of BP elements in messages : Some works supposed that BP elements are explicitly mentioned in messages (e.g., Van der Aalst et al 30 assume that BP names are explicitly included in the email subject). This is actually not realistic as messages senders do not necessary include such information explicitly in messages in a regular way and with unique tags (see Section 3.2.1).…”

“…A4: The lowest granularity expression level of speech acts and activities in messages : Existing approaches have considered a message, 6,7,14,25,26,32‐34 an message subject 30 or a message sentence 16‐21,25 as the lowest structural level that would express: (1) one activity, and/or (2) one sender speech act. In the context of non‐controlled textual data as messages, the expression of these two elements would not be constrained by messages' structure or punctuation (Section 3.2.2).…”