Purpose: The purpose of this project is to improve student word learning by developing an intelligent computer system that annotates text and provides related pictorial information (i.e., pictures, diagrams). This may benefit all students, but especially those who are struggling, by providing explicit scaffolding for word learning in context through multimedia, multilingual, "in the moment," vocabulary assistance. The system will focus on content areas such as science and social science, which are taught in whole-group settings. When reading science and social studies texts, students often encounter novel words, and those with impoverished vocabularies may face difficulty in learning new words.

Project Activities: The project will develop an intelligent computer system that will provide explicit scaffolding for word learning during reading. The research team plans to leverage artificial intelligence algorithms that will automatically generate high-quality vocabulary annotations in English or the reader's native language, along with pictures and pronunciation keys. Feedback will be individualized based on estimates of the reader's ability, thereby providing appropriate levels of information. The research team will conduct a pilot study that will address the feasibility of implementing the intervention in authentic education delivery settings as well as the promise of the intervention for generating outcomes the intervention is designed to effect.

Products: The expected products of this project are an adaptive, intelligent tutor that assists middle-school readers in learning meanings of new words, and published reports.

Structured Abstract

Setting: Participating middle schools in California.

Population: Participants in this project include approximately 80–90 struggling readers attending middle schools with substantial numbers of English language learners and/or students receiving free and reduced price lunches.

Intervention: The computer-assisted learning system will automatically annotate a large number of words and consider the context in which it appears, and so will be able to create different annotations for the same word based on context. Using algorithms, the system will determine which words are likely unknown to the reader and automatically generate appropriate pictures, pronunciation keys, and dictionary entries. The system continually assesses the interaction with the student and modifies its instruction as the reader engages it (and improves). The feedback will be individualized to each student by considering his/her background characteristics, language level (e.g., as assessed by performance on the California Standards Test), and native language. Individual differences in student vocabulary knowledge will be registered in particular lexicons, allowing the personalized system to target unknown words for each student rather than using the broadband approach typically found in traditional instruction. Teachers will be integrally involved in the instruction and development of the annotation. They, along with parents, will actively contribute to "social annotation" through Web 2.0 techniques to further improve the quality of future annotations. The system will generate a report to the teacher to summarize how well the whole class, and how well each individual student is doing, thus providing instructional feedback for all aspects of the curriculum.

Research Design and Methods: The research team will develop the intervention using small focus groups, along with input from teachers with an interest in technology who serve English language learners in either English language development or science classes. A pilot study will be conducted that addresses the feasibility of implementing the intervention in authentic education delivery settings, as well as the promise of the intervention for generating outcomes the intervention is designed to effect.

Key Measures: A number of dependent measures will be examined, such as the patterns of access for the annotations, the end-of-unit scores (analyzed for overall correctness, vocabulary knowledge, and ability to explain concepts), scores on the Word Rating Guide, and scores on the Cloze reading test.

Data Analytic Strategy: In the pilot study, performance will be assessed by conducting a mixed-model 2-way analysis of variance. Students' gain (post-test minus pre-test) scores will be the within-subject factor and the order of exposure to the intervention, which will be counterbalanced by classroom, will be the between-subjects factor.