Assisting students’ writing with computer-based concept map feedback: A validation study of the CohViz feedback system

Abstract: CohViz is a feedback system that provides students with concept maps as feedback on the cohesion of their writing. Although previous studies demonstrated the effectiveness of CohViz, the accuracy of CohViz remains unclear. Thus, we conducted two comprehensive validation studies to assess the accuracy of CohViz in terms of its reliability and validity. In a reliability study, we compared the concept maps generated by CohViz with concept maps generated by four human expert raters based on a text corpus comprising students' explanatory texts (N = 100). Regarding the depiction of cohesion gaps, we obtained high accordance between the CohViz concept maps and the concept maps generated by the human expert raters. However, CohViz tended to overestimate the number of relations within the concept maps. In a validity study, we examined the validity of CohViz and compared central features of the CohViz concept maps with convergent linguistic features and divergent linguistic features based on a Wikipedia text corpus (N = 1020). We found medium to high agreement with the convergent cohesion features and low agreement with the divergent features. Together, these findings suggest that CohViz can be regarded as an accurate feedback system to provide feedback on the cohesion of students' writing

Standort
Deutsche Nationalbibliothek Frankfurt am Main
Umfang
Online-Ressource
Sprache
Englisch
Anmerkungen
PLOS ONE. - 15, 6 (2020) , e0235209, ISSN: 1932-6203

Klassifikation
Literatur, Rhetorik, Literaturwissenschaft

Ereignis
Veröffentlichung
(wo)
Freiburg
(wer)
Universität
(wann)
2020
Urheber

DOI
10.1371/journal.pone.0235209
URN
urn:nbn:de:bsz:25-freidok-1665963
Rechteinformation
Der Zugriff auf das Objekt ist unbeschränkt möglich.
Letzte Aktualisierung
14.08.2025, 10:44 MESZ

Datenpartner

Dieses Objekt wird bereitgestellt von:
Deutsche Nationalbibliothek. Bei Fragen zum Objekt wenden Sie sich bitte an den Datenpartner.

Beteiligte

Entstanden

  • 2020

Ähnliche Objekte (12)