|Title:||Using Computational Linguistics to Detect Comprehension Processes in Constructed Responses across Multiple Large Data Sets|
|Principal Investigator:||McNamara, Danielle||Awardee:||Arizona State University|
|Program:||Cognition and Student Learning [Program Details]|
|Award Period:||2 years (07/01/2019 - 06/30/2021)||Award Amount:||$600,000|
Co-Principal Investigators: Magliano, Joseph; Allen, Laura; O`Rourke, Holly; McCarthy, Kathryn
Purpose: The purpose of this project is to better understand student coherence building, which students use to develop a coherent mental model of a text. Understanding text is a vital activity, enabling us to fully engage in our communities, whether through printed advertisements, electronic messaging, or signs for highway navigation. Students who struggle with coherence building have difficulty achieving deep comprehension. Using previously collected data sets, researchers will examine coherence-building strategies and explore the moderating effects of individual differences across multiple constructed response tasks and texts.
Project Activities: The research team will code and analyze studentsí constructed responses generated from prompts to think aloud and self-explain while reading. The data sets were collected in studies carried out in face-to-face contexts as well as automated reading strategy intelligent tutoring systems. Analyses will focus on identifying indicators of coherence-building and establishing their relations to text comprehension, individual differences, and task constraints.
Products: Researchers will provide evidence of how coherence-building supports critical aspects of text comprehension and how individual differences and tasks moderate these processes. They will also produce peer-reviewed publications and presentations.
Setting: The data sets are from studies conducted in urban and suburban high schools and universities in Arizona and Illinois.
Sample: The data sets include 791 high school students and 1,111 college students from two- and four-year institutions, including students who have been designated as struggling readers through college admissions standards.
Malleable Factor: The malleable factor of interest is studentsí coherence-building strategies and processes during reading.
Research Design and Methods: The research team will analyze data sets collected through multiple prior studies, in which students were asked to respond to prompts, either to think-aloud or self-explain, while reading. The team will examine the moderating effects of individual differences across multiple constructed response tasks and texts. Data analyses will incorporate measures of studentsí skills and motivation collected during the previously conducted studies, computational linguistic analyses of studentsí constructed responses, and expert judgments of comprehension strategy use. In addition, researchers will conduct replication analyses across the multiple data sets to examine the reproducibility of the outcomes.
Control Condition: Due to the nature of this project, there is no control condition.
Key Measures: Key measures include studentsí constructed responses to think aloud protocols and self-explanation prompts, vocabulary knowledge, reading skills, working memory, prior knowledge, metacognition, and motivation.
Data Analytic Strategy: The research team will use multiple linear regression models and linear mixed-effects models to explore how comprehension depends on linguistic features of constructed responses while controlling for individual differences. They will also incorporate analyses from dynamic systems theory to understand how readers coordinate the language in their constructed responses with the language in the text. These models allow the team to quantify stability and change in the properties of constructed responses. Finally, the research team will use machine learning techniques to develop algorithms that predict coherence-building processes and comprehension performance.
Related IES Projects:
Assessing Reading Comprehension with Verbal Protocols and Latent Semantic Analysis (R305G040055)
Project Website: www.distributedliteracy.org