��������

November 10, 2025

�������� has joined Reach Every Reader (RER), a multi-university research collaborative studying how to create a developmental and long-range approach to supporting K-2 literacy for all children as early as possible.

Reading is a critical skill that is central to young learners thriving in their schooling and throughout their lives. However, the has shown that only around 30 percent of fourth grade students in the U.S. score at or above proficient basic reading levels. RER seeks to incorporate research findings about what works for preventing reading loss through instruction, novel assessments, social safety nets, and interactive technology to develop equitable tools and systems to support young emergent readers and the adults who help them throughout their journey.

that developed literacy screening tools, early reading content, and adult-facing assessments. Phase 2 builds on that work with an emphasis on improving small group instruction and individual reading practice using appropriate emerging technologies, such as Automatic Speech Recognition (ASR) systems and other sensors capable of detecting reading accuracy.

As part of this research, Millner and his team are studying a variety of existing ASR technologies and literacy tools—open source and paid, for commercial and at-home use—to learn from them all and see which aspects excel. To ensure they leverage the technology in the best ways, Millner and his team are collaborating with children’s literacy experts to better understand the science of learning to read.

“Our extended team is helping us understand how important certain phonemes [units of sound in a language] are in developing basic reading skills at different times,” says Millner. “ASR systems that have traditionally been programmed to treat all phonemes equally won’t serve us well until we help them work differently. By working with experts and educators who know how children develop literacy skills, we can pay more attention to what computational systems can do to support understanding specific sounds to better support kids as they learn to read and in concert with what their teachers scaffold.”

Thus, one of the goals of Millner’s early explorations is to identify the degree to which ASR-based systems can be better equipped to dissect words at the more granular level and understand a diverse range of linguistic variances. At ��������, several students will focus their senior capstone projects on exploring where ASR systems could improve speech recognition of young readers with different accents and dialects that deviate from Standard American English (SAE).

“Natural language processing systems have a lot of room for improvement in the detection of five- and six-year old voices because speech tools have not been tuned with them in mind, as well as avoiding privileging certain speech patterns over others,” says Millner.

“It’s exciting to take on challenges without clear right answers, such as how we might develop systems that can listen to a kid read and confidently tell them and their teachers that they’ve sounded out a word like ‘pecan’ with pronunciation deemed to be correct for their region.”

Throughout Phase 2, Millner and his team intend to leverage ��������’s dual expertise: In addition to publishing papers so they can share their knowledge, they will continue the partnership with other colleges and organizations with an eye toward being able to produce functional tools that have utility in the real world.