The Intelligent Reading Specialist
Powered by the Science of Reading
SoundScout is LUCA's phoneme-level speech recognition engine that identifies which specific sounds cause reading errors — not just whether a word was right or wrong. Now you can finally know exactly what to work on.
What SoundScout Hears
Child reads:
/s/ /ɪ/ /p/
"sip"
Text expects:
/ʃ/ /ɪ/ /p/
"ship"
/sh/ to /s/ substitution detected: digraph weakness
Trusted by schools and backed by leading research institutions
SoundScout is LUCA's phoneme-level speech recognition engine that identifies which specific sounds cause reading errors, not just whether a word was right or wrong. While most reading tools document failure after the fact, SoundScout gives parents and educators the precise signal needed to fix the right thing the first time. Powered by LUCALabs, SoundScout is the Listen phase of LUCA's three-part intelligent reading cycle.
“We practice reading every night, but my child is still missing the same sounds. I don't know which ones to focus on.”
“The reading app just says 'incorrect.' It never tells me why my child is getting it wrong.”
“We've tried three different programs. They all track progress, but none of them explain what's actually causing the problem.”
“My child is practicing, but I worry they're drilling sounds they already know while the real gap goes unnoticed.”
Word-level tools document failure. They don't prevent it.
When a reading program only marks words right or wrong, it gives you a report card with no road map. Parents spend time on the wrong things. Children practice skills they've already mastered while the real gap stays hidden.
See how LUCA fixes this
SoundScout hears every sound your child makes at the phoneme level, the same precision a specialist uses in a one-on-one session, so LUCA knows exactly where to focus next.
Phoneme-level precision
SoundScout captures every sound your child produces during oral reading and segments it into individual phonemes, timestamped and isolated.
Pattern recognition across sessions
Each flagged sound is matched against LUCADictionary's 763,000+ grapheme-phoneme mappings. Mismatches build into a pattern map that shows exactly which skills are missing.
Targeted practice, not busywork
The signal drives StoryGen and Assessment Intelligence to create stories and sessions that drill the exact sounds causing difficulty, and skip everything your child already knows.
Here is what changes when LUCA listens at the sound level instead of the word level.
Without Phoneme-Level Listening
Student reads "sip" instead of "ship"
Result: Marked incorrect. No further detail.
Parent sees: 1 error on this word
No information about which sound caused the mistake.
Next step: General review
Practice may target skills your child has already mastered. The real gap stays hidden.
With SoundScout
Student reads "sip" instead of "ship"
SoundScout identifies a /sh/ to /s/ substitution at the onset position.
Pattern recognized: Digraph weakness
This is the third /sh/ substitution in 200 words, confirming a systematic gap.
Next step: Targeted /sh/ practice
StoryGen builds stories emphasizing /sh/ words at your child's reading level. So you finally know exactly what to work on, no more guessing.
40% more errors detected compared to word-level recognition (TO VERIFY: link to /our-technology/evidence for published evidence row).
So you finally know exactly what to work on. No more watching your child practice the wrong thing night after night.
Every oral reading session moves through three steps, from raw audio to actionable signal.
Audio captured during oral reading is segmented into individual phonemes using advanced speech recognition. Each sound your child produces is isolated and timestamped.
Each spoken phoneme is matched against the expected phonemes from the text using 763,000+ grapheme-phoneme mappings in LUCADictionary. Mismatches between spoken and expected sounds are flagged with their position and type.
Detected errors are classified by type (substitution, omission, insertion, vowel confusion) and aggregated into patterns. This signal is sent to Assessment Intelligence to identify systematic weaknesses and drive targeted practice in StoryGen.
Each error type reveals a different underlying skill gap, enabling precise intervention.
| Error Type | Example | What It Reveals |
|---|---|---|
| Substitution | "sip" for "ship" | /sh/ to /s/ confusion (digraph weakness) |
| Omission | "bes" for "best" | Final consonant blend difficulty |
| Insertion | "buhlack" for "black" | Blend separation habit |
| Vowel Confusion | "bit" for "bet" | Short /e/ and short /i/ confusion |
Pilot results. See full evidence for methodology and study details.
SoundScout powers the Listen phase, the entry point for every reading session. Its signal flows into Assessment Intelligence and StoryGen to create a practice cycle that gets more precise with every session.
Captures every sound at the phoneme level during oral reading.
Maps phoneme errors into skill patterns and identifies systematic gaps.
Generates targeted stories and sessions that drill only the sounds causing difficulty.
NSF SBIR Grant Recipient
Federal research grant. Only 3% of applicants selected.
Carnegie Mellon University
Research partnership in language and learning technology.
U.S. Patent No. 12,394,332 B2
Proprietary phoneme-level recognition methodology.
NewSchools Venture Fund
Selected for impact-driven education innovation.
SoundScout captures audio during oral reading and segments it into individual phonemes. Each spoken phoneme is aligned against expected phonemes from the text using LUCADictionary's 763,000+ grapheme-phoneme mappings, with mismatches classified by type.
Word-level recognition knows if a word was correct or incorrect. Phoneme-level recognition knows WHY it was incorrect, identifying which specific sound caused the error (for example, a vowel substitution or a skipped blend). That difference means your child gets practice targeting the actual gap, not a general review.
Most reading programs use word-level speech recognition that marks words right or wrong. SoundScout identifies specific sound patterns causing difficulty, enabling targeted intervention through JourneyBuilder and StoryGen. Think of it as the difference between a doctor who says 'something is wrong' and one who runs a full diagnostic.
Yes. SoundScout is designed to catch errors that word-level recognition misses entirely, like /th/ to /f/ substitutions or dropped final consonant blends. These are precisely the errors that accumulate into reading gaps when left undetected.
Yes. LUCADictionary includes pronunciation variants accounting for regional differences, ensuring LUCA distinguishes between accent variations and actual decoding errors so your child is never penalized for how they naturally speak.
SoundScout is well-suited for readers who are struggling, as it pinpoints the exact phoneme patterns causing difficulty. LUCA does not diagnose reading disabilities or learning disorders. If you have concerns about your child's reading development, we encourage you to speak with a qualified specialist as well.
See how LUCA listens at the sound level and builds practice around exactly what your child needs.
