"Mining Knowledge Graphs From Loosely Structured Processes..." Prof. Walid Gaaloul (ICSBT 2022)

Keynote Title: Mining Knowledge Graphs From Loosely Structured Processes: A Use Case From Emailing Systems

Keynote Lecturer: Walid Gaaloul
Presented on: 15/07/2022, Lisbon, Portugal

Abstract: Process-oriented data analysis techniques allow organizations to understand how their processes operate, where modifications are needed and where enhancements are possible. These techniques assume that process data have a structured format. This implicitly means that the underlying processes are structured and therefore are totally executed in the information system. However, in several domains, processes are executed outside the information system using informal methods such as communication tools (e.g. email exchange, IM, etc.). Managing complaints from customers, resolving an insurance claim and internal audits are all examples of processes that require ad-hoc day-to-day tasks (e.g., send an email, schedule meetings, record notes, gather feedback, etc.). These processes are unstructured in nature meaning they have a start, but the activities are not necessarily consistent. Furthermore, these processes often involve a lot of people and communication. Analyzing unstructured processes is challenging for two reasons. First, because of the unstructured nature, the corresponding data exhibit a high variability, which can only be explained when associated to a context. This latter is often unpredictable and therefore, it is difficult to store data according to a predefined schema. Second, data of unstructured processes often reside in unstructured sources such as emails. To enable process analysis, automated techniques need to be developed to extract process related knowledge. To address the first limitation, we propose to store process data in knowledge graphs. In particular, we use labeled property graphs, which incorporate RDF as well as any other type of data. Second, we propose automated techniques to construct knowledge graphs by extracting process related data from natural language texts. As a use case, we use emailing systems.

Conference Website: https://icsbt.scitevents.org

Presented at the following Conference: ICSBT, 19th International Conference on Smart Business Technologies