Showing posts in September 2010 Show all posts >
We are unable to detect Flash Player 9 or higher on your system.
(Flash Player 9 or higher is required for this presentation)
Download the most recent version of Flash Player.
Scientific Learning’s Reading Assistant software helps students develop their reading and fluency skills. Have you ever thought about the technology that was used when building this software? When students sit down in front of the computer and begin their session, what is going on “behind the scenes” as Reading Assistant presents students with a passage to read, records their reading and then gives them a quiz at the end of a passage in order to evaluate comprehension of the material? Let’s take a look at how the software is designed.
Reading Assistant software is unique in its ability to listen along and help students as they read out loud and it uses technology to provide a quality Guided Oral Reading experience for students. This guided oral reading practice is crucial to developing reading fluency. Scientific Learning uses a combination of speech recognition technology and knowledge of the reading process to provide this Reading Verification capability.
The Sphinx open source speech recognition system from Carnegie Mellon University is integrated into Reading Assistant and processes the user’s reading. We have enhanced this software to meet the needs of the education market by adding acoustic models for children’s voices and acoustic models for regional dialects. We have also added the capability to adapt to the user’s voice and speaking rate, detect off-task speech, and detect audio issues so that these can be corrected if possible.
Techniques and analysis based on knowledge of the reading task are combined with the core speech recognition system to enable “Reading Verification.” The reading verification enhancements fall into three categories. Timing analysis identifies the hesitations and dysfluent pauses in a student’s reading. Pronunciation error analysis looks for specific mispronunciations, or partial pronunciations, of words. Word categorization allows the system to treat words differently, depending upon their importance in a given text and whether they are new vocabulary. Finally, Reading Verification analysis as a whole guides a user interface designed to promote fluency, by minimizing interruptions and distractions while at the same time providing help when it is needed.
The performance of Reading Verification is optimized using our extensive automated testing capability. Settings, techniques, and acoustic models are tested and adjusted using recorded audio from hundreds of product users. The goal of this optimization is to identify reading errors, but at the same time we must not disrupt fluency. Therefore we do not want to stop a student on an acceptable reading of a word. In the classroom environment, the Reading Verification process must accommodate a wide range of voices (such as different accents) as well as variable audio conditions (including background noise).
Reading Assistant provides essential one-on-one feedback during guided oral reading to develop a student’s reading skills. We use a combination of speech recognition technology and expert knowledge of the reading process to deliver this capability. Our unique ‘Reading Verification’ technology has been awarded three patents so far, with additional applications in process.
Attend one of our popular webinars with thought leaders in learning. Live and pre-recorded webinars are available. Register today!
Categories: Reading Assistant