Using Automatically Computed Type-of-Word Chains in Linear Models to Predict Topic-Relatedness and Quality of Writing

• In Event: 25.066 - Technology for Improving Math, Writing, Collaboration, and Engagement

Abstract

We-Write persuasively is a new intervention designed to improve persuasive writing with upper elementary grade children. The intervention is teacher-led and computer-supported to provide modeling, practice tasks, assessment, feedback, and adaptations designed to promote mastery of the complex persuasive writing skill. An important part of the web-based tool is assessment of writing befitting the quality and standards typically met by teachers. In this research study we present a new tool that has been designed to assess overall writing quality and topic relatedness. The analysis of essays scored by trained raters and the computer tool shows promise.

Authors

• Pablo Nicolai Pirnay-Dummer, Brandenburg Medical School

• Kausalai K. Wijekumar, Texas A&M University - College Station

• Steve Graham, Arizona State University