Sonogram: a next-generation Decibel Analysis machine learning algorithm to help researchers and administrators understand classroom environments through audio recordings.
Mark Baillie, University of Arkansas at Little Rock
Connor Young, University of Arkansas at Little Rock
Thien Duong, University of Arkansas at Little Rock
Jan P Springer, University of Arkansas at Little Rock
Get your classroom audio analyzed for active learning live at this session!
Classroom observations are a valuable but resource intensive data collection method to analyze how faculty professional development actually impacts teaching practices. The Decibel Analysis for Research in Teaching software tool (DART, Owens & Tanner et al 2017) allows researchers to infer the approximate amount of class time educators spent in passive mode (i.e. single voice) or active mode (i.e. multiple voices or no voices). While DART is insightful in large lecture style classes, it has some limitations in smaller classes and discussion based classes.
Our group is developing the next generation DART algorithm called Sonogram that is capable of accurately analyzing recorded audio from various class sizes and with diverse class formats. Sonogram can analyze preprocessed audio, increasing the manageable file types (i.e. zoom recordings). Our audio processing software is also capable of discerning a single voice (e.g., lecture) in contrast to a group discussion where one person is speaking at a time (i.e. single voice interchange) in addition to loud multi-group discussion and silence from students working together. These distinctions allow for a more nuanced view of "single voice" than what was previously possible with DART.
To maximize value to researchers, our software will be able to apply user-entered metadata to provide insight into what proportion of a class has the opportunity to engage in active learning and what percentage of class time is dominated by the instructor's voice. Our aim is to provide finer-grain details of teaching environments for researchers studying the effects of their professional development efforts or administrators trying to understand what type of teaching environments their students are encountering across curricula.
At this session, we will demonstrate the tool in real time by inviting participants to join the research effort with a Sonogram beta.