A Data Mining Toolbox for Collaborative Writing Processes
Year: 2013
Type: dissertation
Abstract: Collaborative writing (CW) is an essential skill in academia and industry. Providing support during the process of CW can be useful not only for achieving better quality documents, but also for improving the CW skills of the writers. In order to properly support collaborative writing, it is essential to understand how ideas and concepts are developed during the writing process, which consists of a series of steps of writing activities. These steps can be considered as sequence patterns comprising both time events and the semantics of the changes made during those steps. Two techniques can be combined to examine those patterns: process mining, which focuses on extracting process-related knowledge from event logs recorded by an information system; and semantic analysis, which focuses on extracting knowledge about what the student wrote or edited. This thesis contributes (i) techniques to automatically extract process models of collaborative writing processes and (ii) visualisations to describe aspects of collaborative writing. These two techniques form a data mining toolbox for collaborative writing by using process mining, probabilistic graphical models, and text mining. First, I created a framework, WriteProc, for investigating collaborative writing processes, integrated with the existing cloud computing writing tools in Google Docs. Secondly, I created new heuristic to extract the semantic nature of text edits that occur in the document revisions and automatically identify the corresponding writing activities. Thirdly, based on sequences of writing activities, I propose methods to discover the writing process models and transitional state diagrams using a process mining algorithm, Heuristics Miner, and Hidden Markov Models, respectively. Finally, I designed three types of visualisations and made contributions to their underlying techniques for analysing writing processes. All components of the toolbox are validated against annotated writing activities of real documents and a synthetic dataset. I also illustrate how the automatically discovered process models and visualisations are used in the process analysis with real documents written by groups of graduate students. I discuss how the analyses can be used to gain further insight into how students work and create their collaborative documents.
Author: Vilaythong Southavilay
Institution: The University of Sydney
Language: English
Subfield: Management Information Systems
Domain: Social Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W265844495
