Speech-Based Learning Analytics for Collaboration | SRI International

Toggle Menu
three young students working on a laptop together

Speech-Based Learning Analytics for Collaboration

SRI is investigating whether and how student speech can be used to assess the quality of group collaboration during interactive learning activities.

Collaboration is a core teaching and learning process, as well as an important 21st century skill that students must be able to master as they progress through school and into their careers. Collaboration is an integral part of science, technology, engineering, and mathematics (STEM) learning. Collaborative learning has been deeply investigated by learning scientists for several decades, and the field has produced robust theory, meta-analyses of overall efficacy, empirical documentation of success factors, and a wealth of designs for interventions using technology. Yet the field has not yet produced ways to rapidly measure the quality of student collaboration, provide teachers with insights about their students’ collaborative learning, and enable teachers to intervene based on the data.

This research project builds on the successful learning analytics paradigm to extend education analytics from individual students to learning analytics for student collaboration. With the observation that much human collaboration is realized through spoken interaction, the project also builds on recent breakthroughs in human language technologies to research and develop automatic, practical, fast analysis of spoken collaborative interaction. Our multidisciplinary team of learning scientists, speech technologists, and teachers will determine the feasibility of automatically analyzing student speech for indicators of specific individual collaboration behaviors as well as of overall group collaboration quality.

We are collecting a corpus of student speech using collaborative learning activities from a research-based mathematics curriculum in carefully controlled acoustic conditions and then in the ecologically valid setting of middle school classrooms. We will produce a hand-annotated corpus to assess the performance of current automatic speech processing technologies for this type of task, and to determine which aspects of the task may pose particular problems for the different speech technologies we will bring to bear.

If successful, this project could lead to advances in learning technologies, including automatically generated information provided to teachers on how well their students are collaborating and automatically generated real-time feedback to students about their group work (with specific scaffolding on how to improve their collaboration). It could also expand the settings and number of students that can effectively participate in collaborative learning and create products to aid collaborative learning that can be used by a wide audience of educators and professionals.

This material is based upon work supported by the National Science Foundation under Award No.DRL-1432606. Any opinions, findings, and conclusions or recommendations expressed in this publication are those of the author and do not not necessarily reflect the views of the National Science Foundation.