Bringing Speech Recognition to Reading Instruction
As everyone who follows such things knows, U.S. students, as a group, do not read very well. Yet, if you are among those who have read about this—indeed, if you are among those who are reading this Commentary, then you (and most of your friends, neighbors, relatives, and colleagues) are very likely a member of that subset of Americans I would term the “hyper-educated.”
By “hyper-educated,” I do not mean extraordinarily highly educated, though many are. By “hyper-educated,” I mean that you accept that becoming educated is part of the fabric of life; you never questioned that your children would be educated, and you raised them accordingly from the start. In fact, most of the children of hyper-educated Americans read quite well; that is good. Not so good, however, is a resulting tendency for too many of the hyper-educated to think of children with reading difficulties as the exception.
To the contrary, among U.S. students, it is good readers who are the exception. As documented yet again by the recently released National Assessment of Educational Progress, or NAEP, report on reading for 2011, only one in three U.S. students is able to read and understand grade-level material. Still worse, this statistic holds across school grades and has barely budged over as many years as NAEP has tracked it. Moreover, the degree of the literacy deficit is tightly correlated with the extent to which children depend on school (as distinct from home) for their formal education. The irony, of course, is that the fundamental mission of public schooling is to offer educational opportunity—including laying the foundation for reading well—to all children, regardless of what their homes might offer.
Toward this end, I have what some would call an unconventional idea for improving American children’s reading skills; specifically, embracing the use of voice-recognition software in our nation’s classrooms. It is a solution that will take the support of the “hyper-educated” so, please, hear me out.
It is not that our schools are performing more poorly than in years gone by, but that they have never been very good at teaching kids to read. Today’s students don’t read worse than those of yesteryear, but they read no better, either. The problem is that, today, the literacy demands for a productive, self-sufficient life have increased dramatically. Both individually and collectively, both socially and economically, the future of our country depends vitally on the education of its people.
Nor is it that we haven’t tried to fix this situation. As a recent example, the goal of the federal Reading First initiative was to make sure that all children would leave the primary grades having securely learned and understood the alphabetic basics. Coming at the problem from the other direction, the Common Core State Standards Initiative is centered on ensuring guidance and practice with more sophisticated and informative texts.
Both of these initiatives are important and well-founded, but there is also a lot that must happen in between the two. For students to grapple productively with the intellectual challenges of complex texts, they must first gain the ability to read with fluency and ongoing comprehension. It is with this intermediate challenge that most of our students fall by the wayside. In view of this, this intermediate reading period is where I chose to concentrate in a report released recently on technology for developing children’s language and literacy. I wrote the report for the Joan Ganz Cooney Center at Sesame Workshop with the support of the William and Flora Hewlett Foundation.
To most, it is obvious that learning to recognize printed words involves skills and practice specific to the written domain. Yet, this is equally true of the vocabulary, grammar, background knowledge, and modes of thought that characterize text. On every dimension, the comprehension requirements of written language are more demanding, less forgiving, and in many ways qualitatively different from those that characterize oral-language situations. And two overarching factors make this situation still tougher: The first is that, because the knowledge and skills required for reading and understanding written language are specific to written language, their acquisition can come about only through experience in reading and understanding written language. The second is that what has not been understood cannot be learned.
It follows that unless and until children can read and understand texts on their own, they need support and instruction to help them through the task. The obvious reason for providing such help is so students gain from the text at hand. The more important reason is so they will be better able to manage the next text on their own.
As I argue in the report, the real crux of the reading problem lies not with the teachers, the parents, the students, television, the Web, or any of the usual culprits to which blame is often passed. The problem instead is that the individual support required for helping children learn to read is way beyond the capacity of the traditional classroom. Children learn remarkably quickly given the opportunity, but again, one cannot learn what one does not understand. No matter how she tries, the classroom teacher cannot give each of her 20 or so students the individual support on which learning to read depends.
With this issue in mind, the specific recommendation in the report is that our country get serious about developing speech-recognition-based reading software for our schools. This is not a pie-in-the-sky proposal. Today, people around the world, using dozens of languages, depend on automatic speech recognition for telephone call-routing and directory assistance. It is widely employed in dictation and information capture in the defense, heath-care, and legal sectors. It is used for captioning live television so we can watch our favorite games in noisy sports bars, and by unnamed agencies for transcribing suspicious communications. It is used by people to talk to their computers and mobile devices, for example, while browsing the Web, creating voice commands, and managing their bookmarks. People use automatic speech recognition to issue commands to their cell phones and, in reverse, to ask their cell phones to transcribe their voice mail and send written copies to their email. They also use it to talk to their TVs, their music players, their cars, and their navigation systems. And, of course, speech recognition is very hot in the gaming industry.
In other words, automatic speech recognition is a technology that is mature and even commonplace in industry after industry, with the salient exception of where it is needed most: education. Whatever the economic or social value of the applications mentioned above, most pale in comparison to the potential of speech recognition as it could and should be used to help people learn to read and read to learn.
Given “ears,” the computer can listen to students as they read, offering help or prompting further thought at just the right moments, while making records of their progress and difficulties in the background. Such technology, in other words, could provide the individualized, one-on-one, interactive support and guidance on which becoming a reader so integrally depends.
In their potential for providing ample, affordable, effective reading support to every child, I believe that speech recognition-based reading applications should be a priority. Were we to redirect just a fraction of the time, genius, and creativity now devoted to developing ever more seductive ways for us to play games, to watch unwelcome ads, and otherwise to waste our time with our mobile devices and computers, we could do this. But until we somehow convince the hyper-educated to support such innovation, it will not happen.
Vol. 31, Issue 13