ISBN 978-3-8439-1954-8

978-3-8439-1954-8, Reihe Informatik

Kristin Tuot
Process-Driven Document Analysis and Understanding

330 Seiten, Dissertation Technische Universität Kaiserslautern (2014), Hardcover, B5

Zusammenfassung / Abstract

With the shift from traditional to electronic communication, enterprises must manage an increasing number of customer requests arriving through traditional and new communication channels. Existing solutions, in theory as well as in praxis, are insufficient to support efficient and effective request processing over all input channels. Current input channel management reveals two major shortcomings: (1) the disconnect between communication process and internal business processes (BPs) and (2) the fragmentation of existing input channel management systems. To address these shortcomings, a multichannel management approach is needed that allows the automated extraction of information from input channel documents and that integrates multiple input channels simultaneously.

This work aims at providing and evaluating such a holistic approach for enterprises’ multichannel management at the intersection of three research fields, including input channel management, BP integration and, document analysis and understanding (DAU). The approach of process-driven DAU is proposed, wherein internal process information expectations toward incoming documents are formalized to guide DAU. On the BP side, the concept of attentive tasks (ATs) is introduced to formalize the information a process instance expects to be contained in an incoming document. For DAU, all available methods are described in a multichannel Specialist Board (SB) based on a concept introduced by Dengel and Hinkelmann [1996]. This SB allows to dynamically generate and adapt DAU programs for various application domains and multiple communication channels. To make use of these two formalization concepts, the approach iteratively maps an incoming document to one AT from a set of all currently waiting ATs. This AT is used to generate and adapt the DAU plan intended to extract the information expected by the AT.

The process-driven DAU approach is developed sequentially along five major hypotheses in collaboration with two case study partners: (H1) relevance of process-driven DAU, (H2) extraction performance improvements, (H3) efficient and robust mapping of requests to tasks, (H4) continuous planning for DAU performance improvements, and (H5) processing quality and efficiency improvements.