Computer‐Automated Scoring of Written Responses

Abstract

This chapter begins by briefly discussing the human scoring procedures that preceded—and still operate parallel to—computer‐automated scoring (CAS) of written responses. The current conceptualization of the topic is approached by tracing the development of CAS in two areas: extended response tasks such as essays, and limited production tasks such as short answer questions. Limited production responses will be further divided based on the approach to scoring that is being used. This classification is important not only because of the differences in the types of expected responses that they yield, but also because of the different computational approaches normally used to score them: various forms of key word or phrase matching for limited production responses, and systems using more complex forms of natural language processing to score both limited production and extended response tasks. The chapter next moves on to a discussion of current research on CAS in written responses, maintaining its organization based on extended and limited production tasks, and then concludes by exploring the future directions in which research, development, and operational use are likely to proceed.

The Effect of Using Online Automated Feedback on Iraqi EFL Learners’ Writings at University Level

“…They can only be a kind of extra assistance to the instructors but not rather a substitute. So, assessment and grading should be done by instructors rather than by software because writing is a human act, and there is no correlation between how human react and software (Carr, 2014;Koskey & Shermis, 2013;Landauer, 2003). Moreover, other rejections about OAF are that the feedback itself is vague and sometimes incomprehensible to the students (Hegelheimer, 2015;Lai, 2010;Tsuda, 2014).…”

Computer-Based English Language Testing in China: Present and Future

“…Thirdly, more independent, transparent, and comparative research on the quality of automated evaluation engines is needed to assure test takers that they are assessed fairly. All the Chinese empirical studies to date have reported how well their automated evaluation systems predicted human scores; however, as Carr (2014) rightly pointed out, research on automated evaluation systems "has been conducted by the companies developing the systems, and…there is a marked lack of independent research comparing different systems head to head." This limitation also applies to the existing Chinese studies.…”