Duygu BektikPhD Research Student
I joined KMi as a full-time PhD Student in October 2012. My PhD is being supervised by KMI's Simon Buckingham Shum, Anna De Liddo and IET's Rebecca Ferguson.
The main question that this PhD aims to answer is: "To what degree can computational text analysis and visual analytics be used to support the academic writing of students in higher education?"
This research investigates:
1) whether computational techniques can automatically identify the presence or absence of attributes of good academic writing, as correlated with grades and as identified in the literature
2) if this proves possible, how best to feed back actionable analytics to support students and educators
3) and whether this feedback has any demonstrable benefits.
To answer the main research question, the following four supplementary research questions are aimed to be addressed:
RQ1: To what extent is the rhetorical parser XIP accurate and sufficient for identifying the presence of attributes of good academic writing within student writing, as judged by the grade, and by educators?
RQ2: In what ways should XIP output be delivered to end users (students and educators)?
RQ3: To what extent do educators value the results of XIPís analysis of an individual student or cohortís work when the primary focus is on assessment?
RQ4: To what extent do students value the results of XIPís analysis as formative feedback on their writing?
One contribution of this PhD will be a conceptual framework bridging between educational research into academic writing (the features deemed to be important to quality writing), and research into language technologies (the features which can be automatically detected). This framework will occupy the Ďmiddle groundí between learning and computation, helping members of both communities articulate, in precise terms, the opportunities for pedagogically sound learning analytics. This provides the rationale for investigating rhetorical parsers, whose feature extraction capabilities overlap with key features of academic writing.
A second contribution will be the evaluation of a particular rhetorical parser, XIP, against several measures of quality (RQs 1-4), both quantitative and qualitative, engaging both educators, and students from undergraduates to PhDs. Since this is a technology developed outside education, the PhD will exemplify a holistic learning analytics approach to the evaluation of commercial analytics products when migrated into learning contexts, which should be of interest to the wider LAK community given the rapid growth of this market.
Team: Anna De Liddo
Bektik, D., (2014) Visual Analytics of Academic Writing, Demo at The 4th International Learning Analytics and Knowledge, Indianapolis, IN, USA, pp. 265-266, ACM New York, NY, USA ©2014
Bektik, D., (2013) XIP Dashboard: Visual Analytics from Automated Rhetorical Parsing of Scientific Metadiscourse, Workshop: 1st International Workshop on Discourse-Centric Learning Analytics at Learning Analytics and Knowledge (LAK '13), Leuven, Belgium
Taibi, D., (2013) Visualizing the LAK/EDM Literature Using Combined Concept and Rhetorical Sentence Extraction, Demo at Learning Analytics and Knowledge (LAK '13), Leuven, Belgium