Resources
Practical speech technology resources—without the hype.
74 evidence-informed guides for SLPs, clinics and families: home practice, speech sounds, speech AI, multilingual speech, clinical workflow and parent education. Each article is dated and clearly labeled by review status.
Browse
Guides by topic.
How to make articulation home practice easier to track
Home practice becomes easier to track when the assignment is specific, completion is simple to record, and the SLP can see enough context to interpret what happened—not just a binary “done”.
Draft · 5 min read · Drafted Apr 2026Home Practice & CarryoverWhy speech practice between sessions is hard to measure
Between-session speech practice is hard to measure because completion, production quality, cueing, recording conditions and family context are different variables—and most traditional homework captures only one of them.
Draft · 5 min read · Drafted Apr 2026Home Practice & CarryoverWhat “carryover” means in speech therapy
Carryover describes the transfer of a learned speech skill from structured practice into less supported contexts such as phrases, conversation and everyday communication; it is not the same as succeeding on one drill.
Draft · 5 min read · Drafted May 2026Home Practice & CarryoverHome practice vs therapy: what each one is for
Therapy and home practice are complementary, not interchangeable: the SLP evaluates, selects targets and adapts treatment, while home practice can provide additional opportunities to rehearse the clinician-defined goal.
Draft · 5 min read · Drafted May 2026Home Practice & CarryoverHow much articulation practice should happen at home?
There is no single evidence-based number of minutes or repetitions that is right for every child, target and treatment approach; the home dose should follow the SLP’s plan and be realistic enough to sustain.
Draft · 5 min read · Drafted Jun 2026Home Practice & CarryoverWhy shorter speech practice can be easier for families to sustain
Shorter practice can reduce friction because it is easier to fit into routines and may lower behavioral and scheduling barriers, but “short” is not automatically effective—the task still needs a clear target and adequate practice opportunities.
Draft · 5 min read · Drafted Apr 2026Home Practice & CarryoverHow SLPs can assign more specific home practice
A strong home-practice assignment states the target sound, word position, practice level, word set, approximate dose, cueing guidance and what the family should do when the child struggles.
Draft · 5 min read · Drafted May 2026Home Practice & CarryoverA parent-friendly articulation home practice checklist
A parent-friendly checklist should make the routine easy to start, easy to stop and easy to report back to the SLP without asking parents to make specialist clinical judgments.
Draft · 5 min read · Drafted Jun 2026Home Practice & CarryoverHow to reduce friction in speech homework
Friction falls when home practice requires fewer decisions: one clear target, a predictable routine, short setup, easy recording or completion tracking, and a clear path when something is not working.
Draft · 5 min read · Drafted Aug 2026Home Practice & CarryoverWhat information should come back to the SLP after home practice?
The most useful return data are usually the assignment, practice days, number of attempts, recording-quality issues, cue/support context, uncertain examples and any clinician-confirmed labels—not a single opaque “accuracy” score.
Draft · 5 min read · Drafted Jun 2026Home Practice & CarryoverHow to support practice without correcting every word
Parents can support speech practice by following the SLP’s assigned routine, keeping the interaction positive and letting the clinician-defined feedback system handle scoring rather than correcting every spontaneous word.
Draft · 5 min read · Drafted Jul 2026Home Practice & CarryoverWhy practice adherence and speech accuracy are different metrics
Adherence describes whether and how consistently practice occurred; accuracy describes how a target was produced under defined conditions. One can improve without the other, so they should never be merged into a single progress number.
Draft · 5 min read · Drafted Apr 2026Speech SoundsPhoneme-level speech analysis explained
Phoneme-level analysis looks below the word transcript to examine the individual speech-sound sequence and where a production may differ from the expected target.
Draft · 5 min read · Drafted Apr 2026Speech SoundsWhy word recognition is not the same as articulation analysis
Word recognition asks “what word was intended?”; articulation analysis asks “how was the target sound produced?” Those goals can produce opposite system behavior when a recognizer normalizes a misarticulation into the intended word.
Draft · 5 min read · Drafted Jun 2026Speech SoundsWhat is a speech sound substitution?
A substitution occurs when one speech sound is produced in place of another, such as /w/ for /ɹ/ in an English example; whether a pattern is developmentally expected, dialectal or clinically significant depends on context.
Draft · 5 min read · Drafted Jul 2026Speech SoundsSubstitution vs omission vs distortion
A substitution replaces a target sound, an omission leaves a sound out, and a distortion changes the quality of a sound without necessarily turning it into another clear phoneme.
Draft · 5 min read · Drafted Apr 2026Speech SoundsInitial, medial and final word position explained
Word position describes where a target sound occurs in a word. A child may produce the same phoneme differently at the beginning, middle or end because surrounding sounds and syllable structure change the task.
Draft · 5 min read · Drafted Aug 2026Speech SoundsWhat does /r/ carryover mean?
/r/ carryover means an /r/ production learned in structured tasks is appearing more consistently in less supported speech, eventually including connected conversation—not merely that isolated /r/ can be produced.
Draft · 4 min read · Drafted May 2026Speech SoundsFrom /r/ isolation to conversation: the practice hierarchy
A common practice hierarchy moves from simpler elicitation contexts toward syllables, words, phrases, sentences and connected speech, but the exact sequence and cueing strategy should be chosen by the treating SLP.
Draft · 5 min read · Drafted May 2026Speech SoundsPracticing /s/ across word positions
Practicing /s/ across positions can reveal that initial, medial and final /s/ are not identical tasks; an SLP may intentionally narrow practice to one position before broadening contexts.
Draft · 4 min read · Drafted May 2026Speech SoundsPracticing /l/ from words to connected speech
Moving /l/ from words into phrases and connected speech increases linguistic and motor demands, so performance should be tracked by context rather than collapsed into one global “/l/ score.”
Draft · 4 min read · Drafted May 2026Speech SoundsMinimal pairs: what they are and when SLPs use them
Minimal pairs are words that differ by one meaningful sound contrast and can be used in phonological contrast therapy when that contrast matches the child’s error pattern.
Draft · 5 min read · Drafted May 2026Speech SoundsWhat is Percent Consonants Correct?
Percent Consonants Correct (PCC) is a quantitative measure calculated from the number of correctly produced consonants divided by the total consonant opportunities in a speech sample, using a defined scoring method.
Draft · 5 min read · Drafted May 2026Speech SoundsWhy one percentage cannot describe every speech production
A single percentage can summarize performance but it can hide which sounds, positions, cue levels and error types changed; useful clinical data should let the reviewer move from the aggregate back to the evidence.
Draft · 5 min read · Drafted Jun 2026Speech SoundsWhat “stimulability” means
Stimulability refers to whether a person can accurately imitate a target speech sound when given a model and, sometimes, cues; SLPs may use this information when selecting targets and planning intervention.
Draft · 4 min read · Drafted Aug 2026Speech SoundsSpeech sound acquisition norms: why language matters
Speech-sound acquisition norms are language-specific descriptions of when groups of children tend to acquire sounds; they are useful context but should not be treated as universal deadlines or transplanted from English to another language.
Draft · 5 min read · Drafted Apr 2026AI for SLPsAI in speech therapy: what should remain under clinician control?
AI can assist with repetitive analysis, organization and drafting, but assessment, diagnosis, goal selection, treatment changes and clinical interpretation should remain under qualified clinician control.
Draft · 5 min read · Drafted Jun 2026AI for SLPsHow AI can support articulation home practice
AI can make home practice easier to structure by checking recording quality, organizing attempts, surfacing supported sound-level patterns and routing uncertainty for review; it should not independently prescribe treatment.
Draft · 5 min read · Drafted Jul 2026AI for SLPsWhy generic speech recognition struggles with speech sound errors
General ASR is optimized to recover intended words, while speech-sound analysis needs the model to preserve the exact deviations that ordinary recognition may correct away.
Draft · 5 min read · Drafted Aug 2026AI for SLPsWhat is speech-to-phoneme?
Speech-to-phoneme converts audio into a sequence of phonetic or phonemic units rather than ordinary spelling, creating a representation that can be compared with a target pronunciation.
Draft · 5 min read · Drafted Jun 2026AI for SLPsAutomatic speech recognition vs phoneme analysis
ASR measures how well a system recovers words or text; phoneme analysis focuses on sound-level realization. A product can be excellent at one and poor at the other.
Draft · 5 min read · Drafted Jun 2026AI for SLPsWhy confidence scores matter in clinical speech AI
A confidence score is useful only if it is calibrated to an action: accept, ask for a retry, or send to review. An impressive-looking number without validation can create false certainty.
Draft · 5 min read · Drafted Jul 2026AI for SLPsWhy “I don't know” is an important AI output
In clinical-adjacent speech AI, abstaining can be safer than forcing every recording into a label. “Unable to judge” or “needs review” should be first-class outputs, especially with poor audio or out-of-scope speech.
Draft · 5 min read · Drafted Aug 2026AI for SLPsHuman-in-the-loop speech AI explained
Human-in-the-loop speech AI routes model output through a defined review workflow so a clinician can confirm, correct, exclude or defer findings and the final record clearly distinguishes model output from clinician judgment.
Draft · 5 min read · Drafted Jul 2026AI for SLPsHow to evaluate an AI speech therapy tool
Evaluate the intended use, target population, validation data, uncertainty behavior, clinician controls, privacy practices and update policy—not the demo alone.
Draft · 5 min read · Drafted Apr 2026AI for SLPs10 questions SLPs should ask an AI vendor
The best vendor questions uncover scope, evidence, uncertainty, data handling and control: who was tested, what counts as correct, how the model fails, who reviews it and what happens to recordings.
Draft · 5 min read · Drafted Jul 2026AI for SLPsWhat happens when child speech recognition gets a word wrong?
A recognition error can come from the child’s production, the microphone, background noise, language/dialect mismatch or the model itself; a clinical workflow should not treat all ASR errors as evidence about the child.
Draft · 5 min read · Drafted Aug 2026AI for SLPsWhy low-confidence recordings should go to review
Low-confidence recordings belong in a review or retry path because converting uncertainty into a definite child-facing score can create misleading feedback and contaminate progress data.
Draft · 5 min read · Drafted Jun 2026AI for SLPsHow SLP feedback can improve speech AI safely
Clinician corrections can become valuable labeled data only when the product separates patient care from model-improvement use, preserves provenance and obtains the required consent for secondary training use.
Draft · 5 min read · Drafted May 2026AI for SLPsShould AI-generated SOAP notes be reviewed?
Yes. AI-generated SOAP or progress-note drafts should be treated as drafts: clinicians should verify facts, omissions, certainty, speaker attribution and clinical meaning before signing or using them downstream.
Draft · 5 min read · Drafted Jul 2026AI for SLPsClinical AI benchmarks: what SLPs should look for
A useful benchmark states the population, language, task, reference standard, sample size, model version, uncertainty behavior and subgroup results—not just one accuracy number.
Draft · 5 min read · Drafted Jul 2026Multilingual SpeechWhy an English pronunciation model cannot simply be translated
Translating interface text does not localize a speech model: languages differ in phoneme inventories, allowable sound sequences, syllable structures, developmental expectations and dialect patterns.
Draft · 5 min read · Drafted Jul 2026Multilingual SpeechWhat makes multilingual speech assessment difficult?
Multilingual assessment is difficult because skills are distributed across languages, exposure changes over time, cross-linguistic transfer is normal and monolingual norms can misclassify difference as disorder.
Draft · 5 min read · Drafted Apr 2026Multilingual SpeechDialect difference vs speech sound error
A dialect difference follows the rules of a speech community; a speech-sound error is interpreted relative to the person’s linguistic system. Treating one prestige dialect as the universal target creates avoidable misclassification.
Draft · 5 min read · Drafted Jul 2026Multilingual SpeechWhy phoneme inventories matter in speech AI
A phoneme inventory defines the contrastive sound categories a language uses; a speech-AI system needs the correct inventory and allophonic/dialect context before it can meaningfully compare target and observed productions.
Draft · 5 min read · Drafted Jun 2026Multilingual SpeechBuilding speech AI for Turkish children
Building child-speech AI for Turkish requires Turkish-specific phoneme/phonotactic knowledge, child recordings across ages, word-position coverage and SLP-labeled validation rather than an English model with Turkish text added.
Draft · 5 min read · Drafted Apr 2026Multilingual SpeechTurkish articulation: why word position matters
Turkish articulation assessment samples sounds across positions because phonotactic constraints and surrounding segments affect where and how a sound can appear; “the child can say /r/” is less informative than position-specific evidence.
Draft · 5 min read · Drafted Jun 2026Multilingual SpeechWhy Arabic speech AI needs dialect-aware validation
Arabic speech AI needs dialect-aware validation because Modern Standard Arabic and regional spoken varieties differ in pronunciation, vocabulary and phonological patterns; a model validated on one variety should not claim universal Arabic performance.
Draft · 5 min read · Drafted Jul 2026Multilingual SpeechMultilingual children and speech sound assessment
A multilingual child should not be compared only with monolingual English expectations. Assessment should examine patterns across the child’s languages, exposure history and dialects to distinguish difference from possible disorder.
Draft · 5 min read · Drafted Jul 2026Multilingual SpeechIPA and multilingual speech technology
The International Phonetic Alphabet gives speech technology and clinicians a shared notation for speech sounds, but an IPA symbol is only a representation; automatic transcription still needs language context and validated acoustic models.
Draft · 5 min read · Drafted Jun 2026Multilingual SpeechWhat a speech AI language pack should contain
A responsible speech-AI language pack should include the phoneme inventory, phonotactics, grapheme-to-phoneme rules where relevant, dialect scope, target corpus, developmental references, annotated child data, benchmarks and known limitations.
Draft · 5 min read · Drafted Jun 2026Multilingual SpeechWhy developmental norms must be language-specific
Developmental speech norms describe patterns within specific languages and populations. Applying English ages or elimination patterns to another language can turn normal linguistic differences into apparent delay.
Draft · 5 min read · Drafted Apr 2026Workflow & Clinical OperationsHow to review home-practice data without creating more work
Home-practice data is useful only if the system reduces review burden: summarize routine activity, surface exceptions, preserve evidence and let the clinician drill down when judgment is needed.
Draft · 5 min read · Drafted May 2026Workflow & Clinical OperationsWhat a useful SLP dashboard should show
A useful dashboard should answer what changed, what needs attention and what evidence supports it—without turning the SLP into a full-time data analyst.
Draft · 4 min read · Drafted Jul 2026Workflow & Clinical OperationsHow to organize a speech review queue
A speech review queue should be prioritized by uncertainty, clinical relevance and recording quality, with fast actions for confirm, change label, unable to judge and exclude.
Draft · 5 min read · Drafted Apr 2026Workflow & Clinical OperationsClinical documentation vs speech practice data
Speech-practice data records what happened during assigned attempts; clinical documentation interprets the encounter and professional decision-making. The first can support the second, but should not be treated as the same record.
Draft · 5 min read · Drafted May 2026Workflow & Clinical OperationsWhat should be in a parent progress summary?
A parent summary should emphasize completion, assigned goals, encouraging trends and what the SLP wants the family to do next—without exposing clinical labels or forcing the parent to interpret uncertain AI findings.
Draft · 5 min read · Drafted Jun 2026Workflow & Clinical OperationsHow speech clinics can pilot AI safely
A safe pilot defines the intended workflow, success measures, excluded use cases, data handling, human review and stop criteria before the first patient recording is processed.
Draft · 5 min read · Drafted Jul 2026Workflow & Clinical OperationsA procurement checklist for clinical speech AI
Procurement should ask for intended use, population, model version, validation methods, subgroup/language results, uncertainty handling, data flow, training use, retention, subprocessors, update notice and clinician controls.
Draft · 5 min read · Drafted Apr 2026Workflow & Clinical OperationsWhat to ask about child voice data before buying speech software
Before buying software that records children, ask exactly what audio is collected, why, where it goes, how long it stays, who can access it, whether it trains models and how deletion actually works across backups and vendors.
Draft · 5 min read · Drafted Aug 2026Workflow & Clinical OperationsRaw audio retention: questions clinics should ask
Retention should have a defined purpose and end date. “We keep audio to improve the product” is not a sufficient operational answer for sensitive child speech data.
Draft · 5 min read · Drafted Jun 2026Workflow & Clinical OperationsAI training data and therapy recordings: what is the difference?
Therapy recordings are collected to deliver a service to a patient; training data is reused to improve or build a model. Those are different purposes and should have different consent, access, provenance and retention rules.
Draft · 5 min read · Drafted Jun 2026Workflow & Clinical OperationsHow to design a clinician-in-the-loop AI workflow
Start by defining which decisions are automated, which require review, what evidence the clinician sees, how they override the model and how the system records that decision.
Draft · 5 min read · Drafted Aug 2026Workflow & Clinical OperationsWhen not to automate speech feedback
Do not automate feedback when the model is outside its validated language/population, the recording is not judgeable, the consequence of a wrong message is high, or the output requires a clinical interpretation the system cannot support.
Draft · 5 min read · Drafted Jul 2026Parent EducationMy child is practicing a speech sound at home—what is my role?
Your main role is to make the SLP’s assigned practice easy to complete: create a calm routine, follow the model and cues you were given, encourage effort and send questions back to the clinician.
Draft · 5 min read · Drafted Aug 2026Parent EducationShould I correct my child's speech during home practice?
Correct only in the way your child’s SLP has asked you to during the assigned practice. Constant correction during everyday conversation can create conflict and does not substitute for a clinician-designed cueing plan.
Draft · 5 min read · Drafted May 2026Parent EducationWhat should I do if my child gets frustrated during speech practice?
Pause or shorten the activity, keep the interaction neutral and tell the SLP what happened. Frustration is useful information about the task, timing or support level—it is not a reason to push through at any cost.
Draft · 5 min read · Drafted Jun 2026Parent EducationHow to make speech homework fit into a busy week
Home practice is more sustainable when it is attached to an existing routine, requires little setup and has a realistic stopping point agreed with the SLP.
Draft · 5 min read · Drafted Jun 2026Parent EducationWhy your child's SLP may work on one sound in different word positions
A sound at the beginning, middle and end of a word occurs in different phonetic and syllable contexts, so the SLP may target one position at a time to make the task more specific.
Draft · 5 min read · Drafted Apr 2026Parent EducationWhat does “practice between sessions” mean?
Practice between sessions means rehearsing the specific goals your SLP assigned during the days between appointments; it is not a replacement for the session or an invitation to invent new targets.
Draft · 5 min read · Drafted Aug 2026Parent EducationWhy an app cannot diagnose your child's speech
An app can record, organize or analyze supported speech tasks, but diagnosis requires broader clinical information, trained interpretation and consideration of hearing, language, development, structure, motor speech and context.
Draft · 5 min read · Drafted Jul 2026Parent EducationQuestions to ask before using an AI speech app with a child
Ask who selects the target, what the app has been tested on, how it handles uncertainty, what happens to voice recordings, whether data trains models and how a parent or clinician can correct or delete information.
Draft · 5 min read · Drafted May 2026Parent EducationHow to protect your child's voice recordings in an app
Treat child voice recordings like sensitive personal data: understand the purpose, consent flow, retention period, access controls, training use, deletion process and third-party processors before recording.
Draft · 5 min read · Drafted Aug 2026Parent EducationWhy home practice should follow your SLP's target
Home practice works best as an extension of the clinical plan. Changing sounds, cues or difficulty on your own can make it harder to know what helped and may not fit the child’s speech profile.
Draft · 5 min read · Drafted Apr 2026Publication pipeline: draft → clinical review → published. Articles on this page are complete editorial drafts with sources; they are educational, dated, and not diagnosis or individualized treatment. Status changes only after a qualified reviewer signs off.
Newsletter
Speech AI for SLPs—without the hype.
Product updates, validation notes and practical guides for clinician-guided speech technology. No spam, unsubscribe anytime.