Natural Language Processing for Writing and Speaking

In this chapter we introduce the topic of automated scoring for written and spoken constructed responses and provide a detailed overview of the methodological underpinnings of natural language processing (NLP) that enable it. In the first section, we start with a high-level overview of the types of assessment tasks that automated text and speech scoring systems have been applied to and highlight the relevant aspects of the construct for each task type that motivate the use of automated scoring. In the second section, we provide an introduction to fundamental concepts from the fields of linguistics and NLP that are used in automated scoring technologies. In the third section, we introduce the standard processing pipeline for designing and using an automated scoring system for written or spoken language and provide in-depth overviews of the types of linguistic features that are typically extracted as well as the state-of-the-art NLP and speech processing methodologies that enables this. Finally, we conclude with a discussion of the readiness of automated scoring of text and speech for different use cases and discuss potential future directions for the field.