User Approaches to Knowledge Externalization in Visual Analytics of Unstructured Data

Information

• Publication Type: Master Thesis
• Workgroup(s)/Project(s): not specified
• Date: 2024
• Open Access: yes
• First Supervisor: Manuela Waldner
• Pages: 80
• Keywords: Knowledge Externalization, Knowledge-Assisted Visualization, Visual Analytics, Unstructured Data, Concept Maps, Mental Models, User Study, Data Exploration

Abstract

Traditional machine learning approaches for analyzing large unstructured data often depend on labelled training data and well-defined target definitions. However, these may not be available or feasible when dealing with unknown and unstructured data. It requires human reasoning and domain knowledge to interpret it. Interactive systems that combine human analytical abilities with machine learning techniques can address this limitation. However, to incorporate human knowledge in such systems, we need a better understanding of the semantic information and structures that users observe and expect while exploring unstructured data, as well as how they make their tacit knowledge explicit. This thesis aims to narrow the gap between human cognition and (knowledge-assisted) visual analytics. In a qualitative and exploratory user study, this thesis investigates how individuals explore a large unstructured dataset and which strategies they apply to externalize their mental models. By analyzing users' externalized mental models, we aim to better understand how their knowledge evolves during data exploration. We evaluate the comprehensiveness, detail and evolution of users' external knowledge representations by applying quantitative and qualitative methods, including a crowdsourcing step. The results show that users' externalized structures are able to represent a given dataset comprehensively and to a high degree of detail. While these knowledge representations are highly subjective and show various individual differences, we could identify structural similarities between individuals. In addition to the insights about how users externalize their tacit knowledge during data exploration, we propose design guidelines for (knowledge-assisted) visual analytics systems.

Additional Files and Images

Weblinks

• Entry in reposiTUm (TU Wien Publication Database)
• DOI: 10.34726/hss.2024.115066

BibTeX

@mastersthesis{irendorfer-2024-uat,
  title =      "User Approaches to Knowledge Externalization in Visual
               Analytics of Unstructured Data",
  author =     "Max Irendorfer",
  year =       "2024",
  abstract =   "Traditional machine learning approaches for analyzing large
               unstructured data often depend on labelled training data and
               well-defined target definitions. However, these may not be
               available or feasible when dealing with unknown and
               unstructured data. It requires human reasoning and domain
               knowledge to interpret it. Interactive systems that combine
               human analytical abilities with machine learning techniques
               can address this limitation. However, to incorporate human
               knowledge in such systems, we need a better understanding of
               the semantic information and structures that users observe
               and expect while exploring unstructured data, as well as how
               they make their tacit knowledge explicit. This thesis aims
               to narrow the gap between human cognition and
               (knowledge-assisted) visual analytics. In a qualitative and
               exploratory user study, this thesis investigates how
               individuals explore a large unstructured dataset and which
               strategies they apply to externalize their mental models. By
               analyzing users' externalized mental models, we aim to
               better understand how their knowledge evolves during data
               exploration. We evaluate the comprehensiveness, detail and
               evolution of users' external knowledge representations by
               applying quantitative and qualitative methods, including a
               crowdsourcing step. The results show that users'
               externalized structures are able to represent a given
               dataset comprehensively and to a high degree of detail.
               While these knowledge representations are highly subjective
               and show various individual differences, we could identify
               structural similarities between individuals. In addition to
               the insights about how users externalize their tacit
               knowledge during data exploration, we propose design
               guidelines for (knowledge-assisted) visual analytics
               systems.",
  pages =      "80",
  address =    "Favoritenstrasse 9-11/E193-02, A-1040 Vienna, Austria",
  school =     "Research Unit of Computer Graphics, Institute of Visual
               Computing and Human-Centered Technology, Faculty of
               Informatics, TU Wien",
  keywords =   "Knowledge Externalization, Knowledge-Assisted Visualization,
               Visual Analytics, Unstructured Data, Concept Maps, Mental
               Models, User Study, Data Exploration",
  URL =        "https://www.cg.tuwien.ac.at/research/publications/2024/irendorfer-2024-uat/",
}