Explicit Comprehension Instruction in an Automated Reading Tutor that Listens: Report of Progress, Year One
msra
摘要
How should an automated Reading Tutor that listens and speaks teach reading comprehension strategies? Research on effective classroom reading comprehension instruction suggests key elements to incorporate: (1) use of fictional narrative texts and informational texts well-suited to reading comprehension instruction (e.g., with strong story structure and considerable content); (2) setting a purpose for reading at the outset of each text; and (3) an apprenticeship or gradual release of responsibility model (Pearson & Gallagher, 1983) in which the tutor provides explicit teaching, modeling, and collaborative and independent practice. We are focusing on four reading comprehension strategies supported by research literature and well-suited to technology's affordances constraints: questioning, visualizing, comprehension monitoring, and summarizing. In year 1 we are selecting texts suitable to teach these strategies; scripting the instruction, practice, and assessment to insert; automating these scripts for Project LISTEN's Reading Tutor; user-testing them on second and third graders who use it daily at two elementary schools; iterating on their design based on user-testing; transcribing children's spoken responses recorded by the Reading Tutor; annotating how an expert practitioner would interpret and react to those responses; and using this data to identify useful categories and features of oral responses to detect. This analysis is guiding our efforts to classify spoken responses automatically - initially in off-line experiments on the recorded speech, and subsequently on-line in the Reading Tutor. This work should enable the Reading Tutor -- despite the limited accuracy of speech recognition - to 'listen' for students' strategy use and comprehension processes (1), and respond accordingly.
更多查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要