Fluency for next generation assessments
Looking for a proven and reliable approach to automating oral reading assessments? The SoapBox voice engine was built to deliver accurate and scalable fluency assessments for kids of all ages, accents, and dialects. It works in noisy environments, without the need for headsets, and generates rich and granular voice data to populate your teacher dashboards.
Clients like McGraw Hill and Amplify and partners like MetaMetrics have chosen our speech recognition technology to power their oral reading assessments and have confirmed its performance at human-level accuracy.
How it works
Get under the hood of our Fluency feature and learn about the unique reading assessment data it generates for teachers with our colleague, Computational Linguist Agape Deng.
Accurate down to the phoneme level
This white paper examines the value of voice-powered literacy practice and assessment and introduces a new feature — phoneme-level scoring — which helps educators to pinpoint, down to the phoneme level, where students are struggling and where they would benefit from more personalized instruction and intervention.
Phonemic breakdown in action
Watch how our voice engine returns percentage-based confidence scores for target phrases, words, sentences, paragraphs, and for every speech sound a child utters. Accurate scoring is delivered right down to the phoneme level.
In order to achieve the highest levels of accuracy from our voice engine, customers can request a custom model (CM) tailored to their domain or corpus. CMs ensure that all corpus words, proper nouns, and bespoke words (including names, brand names, experience-specific terms, colloquial terms, and slang) are covered in our dictionaries and models.
To find out more about creating a CM for your use case, let’s start by learning more about them.
FLUENCY DATA FEATURES
WCPM: Data points returned by our voice engine enable users to calculate words correct per minute (WCPM)
Substitutions: When a child says a different word than expected in the reference text
Omissions/Deletions: When a child omits words present in the reference text
Insertions: When a child’s audio file contains extra words not present in the reference text
Repetitions: When a child repeats a word from the reference text (a subtype of insertions)
Pauses: When a child pauses for 1 to 3 seconds while reading
Hesitations: When a child pauses for three seconds or more while reading
Self-corrections: When a child corrects a word they have just read within 3 seconds (new!)
Prosody: The intonation, rhythm, and emphasis given to words and sentences when reading out loud (coming soon)
SoapBox Fluency product sheet
In this product sheet, you’ll learn how voice-enabling your reading product means more classroom assessments, more one-on-one instruction time for teachers, and better outcomes for kids. You’ll also learn about all of the rich data points (reading errors) returned by Fluency to measure accuracy, words correct per minute (WCPM), and more.