Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TimeFlows: Visualizing Process Chronologies from Vast Collections of Heterogeneous Information Objects

Max Lonysa Muller, Erik Saaman, Jan Martijn E. M. van der Werf, Charles Jeurgens, Hajo A. Reijers

TL;DR

TimeFlows addresses the reconstruction of non-repetitive processes from unstructured information objects by introducing a multi-relational visualization that links events to their constitutive documents. Grounded in explorative interviews with domain experts and demonstrated on the Dutch Childcare Benefits Scandal, TimeFlows supports diverse relations (e.g., Temporal, Subject, Entity, Causal) and ties visuals to source objects, enabling perspective-driven understanding. The work outlines a Process Delivery Diagram to model current practice, defines a three-tier relation framework, and proposes a pathway toward automation (NER/NEL, TERNL, EvEx) to scale retrospective process chronology across large, heterogeneous data sources.

Abstract

In many fact-finding investigations, notably parliamentary inquiries, process chronologies are created to reconstruct how a controversial policy or decision came into existence. Current approaches, like timelines, lack the expressiveness to represent the variety of relations in which historic events may link to the overall chronology. This obfuscates the nature of the interdependence among the events, and the texts from which they are distilled. Based on explorative interviews with expert analysts, we propose an extended, rich set of relationships. We describe how these can be visualized as TimeFlows. We provide an example of such a visualization by illustrating the Childcare Benefits Scandal -- an affair that deeply affected Dutch politics in recent years. This work extends the scope of existing process discovery research into the direction of unveiling non-repetitive processes from unstructured information objects.

TimeFlows: Visualizing Process Chronologies from Vast Collections of Heterogeneous Information Objects

TL;DR

TimeFlows addresses the reconstruction of non-repetitive processes from unstructured information objects by introducing a multi-relational visualization that links events to their constitutive documents. Grounded in explorative interviews with domain experts and demonstrated on the Dutch Childcare Benefits Scandal, TimeFlows supports diverse relations (e.g., Temporal, Subject, Entity, Causal) and ties visuals to source objects, enabling perspective-driven understanding. The work outlines a Process Delivery Diagram to model current practice, defines a three-tier relation framework, and proposes a pathway toward automation (NER/NEL, TERNL, EvEx) to scale retrospective process chronology across large, heterogeneous data sources.

Abstract

In many fact-finding investigations, notably parliamentary inquiries, process chronologies are created to reconstruct how a controversial policy or decision came into existence. Current approaches, like timelines, lack the expressiveness to represent the variety of relations in which historic events may link to the overall chronology. This obfuscates the nature of the interdependence among the events, and the texts from which they are distilled. Based on explorative interviews with expert analysts, we propose an extended, rich set of relationships. We describe how these can be visualized as TimeFlows. We provide an example of such a visualization by illustrating the Childcare Benefits Scandal -- an affair that deeply affected Dutch politics in recent years. This work extends the scope of existing process discovery research into the direction of unveiling non-repetitive processes from unstructured information objects.
Paper Structure (13 sections, 4 figures, 1 table)

This paper contains 13 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Research Method
  4. Interviewee Groups
  5. Semi-Structured Interviews
  6. Constructing Process Chronologies
  7. Situationalized Process Chronologies
  8. Identified Relations Used in Document Analysis
  9. Visualizing TimeFlows
  10. TimeFlows to Visualize Process Chronologies
  11. An Illustrative Example
  12. Further Research: Challenges
  13. Conclusion

Figures (4)

  • Figure 1: This US elections TGR liu2020 links events through the Subject Relation.
  • Figure 2: Process Delivery Diagram of constructing TimeFlows based on the participants. The relations $\mathit{TT}$, $\mathit{TE}$ and $\mathit{EE}$ refer to the relations specified in Table \ref{['catgroup']}. Together, these relations form a taxonomy.
  • Figure 3: Number of interviewee groups that mention the respective relationships between events as one of the top three most important ones
  • Figure 4: The Visual Analytics feedback loop visanalytics applied to process chronologies.