For SLPs & clinicians
EZConversation is an AAC app with a grid layout for the iPad. It helps users in creating grammatical sentences, and it uses speech recognition to help users find words that they're likely to say next. This page covers how the grammar works, how to set the app up for a client, and the reporting dashboard.
Go to your dashboard Full feature list
For everything in more detail, read the full SLP manual (PDF).
The design premise
EZConversation does not require the user to encode morphology or syntax. There is no my tile, no and tile, no -ing, no is, and no possessive marker. The user taps content words in roughly semantic order, and the app supplies inflection, articles, copulas, prepositions, conjunctions and case.
This is a trade-off. The benefit is a short path from what the user wants to say to a fluent sentence, for a user who struggles most with function words. The cost is that the user has no direct control over those forms. A user who can mark morphology has no tile with which to do so, and the app's reading of an ambiguous sequence stands unless a partner edits it in the message bar.
It is a good fit for a client with a strong semantic core and weak morphosyntax. It is a poor fit for a client whose goal is independent grammatical encoding, or a literate client who wants word-for-word control. The keyboard's Phrase mode allows typed sentences to be spoken exactly as written, but it is not the main way the app is meant to be used. This should be decided before setting anything up.
What the grammar engine handles (some examples)
The tapped words are split into clauses, and a fixed set of rules is applied to each one. In clinical terms, it covers:
- Case — one tile per person, with accusative and possessive forms chosen by position. "give book I" becomes "give the book to me", and "I like I car" becomes "I like my car".
- Agreement and tense — third-person singular present, past tense from a time word anywhere in the sentence, and future from words like "tomorrow" or "next week".
- Function words — infinitival "to", prepositions a verb requires, and purpose infinitives.
- Articles — "the" and "a/an", left off where English leaves them off (mass nouns, family members, phrases like "go to school").
- Missing "is / am / are" — for adjectives, locations, noun phrases, "there is/are" and wh-questions.
- Negation — a single "no" tile becomes don't, doesn't, isn't, didn't or not, as the clause requires.
- Wh-questions — "where he go" becomes "where does he go", while embedded questions keep statement order ("I know where she goes").
- Coordination — "and" is added where needed, since there is no conjunction tile.
- Body parts — "I hurt knee" becomes "I hurt my knee".
Known gaps
Currently, the app does not construct comparatives and superlatives, dative shift, reflexives, perfect aspect, or prosodic sentence type. Most importantly, past tense comes from a time word. Without one, "I eat" stays in the present tense.
If a sequence of taps consistently produces the wrong sentence, please report it with the exact tap order, what was spoken, and what should have been said.
Setting the app up for a client
- Decide whether the trade-off above fits.
- Personalise before you teach. Add people first, then 5–15 custom words with photos and categories.
- Trim the pinned words to the client's actual core vocabulary. Every pinned word is shown on the Pinned tab, so fewer words make each one easier to find.
- Set access settings (tile size, hold time and repeat-tap delay) and confirm the client can reliably hit their targets before working on language.
- Teach the three tile types to every partner, especially that Instant Words are spoken immediately and are not added to the message bar.
- Set the voice with the client, wherever they can indicate a preference.
- Teach possessives. Tapping a person's tile twice ("I like I car") produces "my", which is not obvious without being shown.
- Introduce Listening Mode last, once the client can navigate topics independently, so it speeds up finding words rather than replacing that skill.
Supporter modeling
When you or a caregiver model language by tapping on the client's own iPad, the app cannot tell those sentences apart from the client's. Without Supporter modeling on, they are logged as the client's and they train the predictive keyboard toward your word choices.
Settings → Supporter modeling prevents both, and shows an orange banner while it is on. It resets off every time the app launches.
If you intend to rely on the reports below, every partner who uses the iPad should know how to turn it on.
The reporting dashboard
The dashboard shows usage data for each client over the last 90 days: total sentences, sentences by week, which topics they came from, the number of unique words, and average sentence length by week.
Linking to a client's account
As a security feature, clinicians must be linked from their client's account.
- Sign in here and choose "I'm an SLP / clinician" as your role.
- On the client's account, sign in and generate an invite code.
- Enter that code under Redeem an invite code on your own dashboard.
Codes last seven days and link one person each. The client's account can see everyone linked to it and remove any of them at any time.
Interpreting the data
- It counts words spoken through the app, not language. It is usage data, and is not a substitute for a language sample.
- Average sentence length counts tiles tapped, not words spoken. Words the app added are not included, and a multi-word tile such as "thank you" counts as one. It is not MLU in morphemes, and should not be reported as MLU without explanation.
- Anything modeled with Supporter modeling off is counted as the client's.
- Data from before the link is included. Once linked, you will see the full 90-day history, including sentences from before you were working with the client.
The reports show counts and word inventories rather than transcripts, though the sentence text is stored on our servers so the reports can be produced. Confirm this fits your setting's records policy before relying on it.
Privacy review
Vocabulary, grammar, topic selection and voice output all run on the iPad. Speech recognition is the one exception: Listening Mode uses the iPad's internal speech recognition, the same one behind dictation, so partner audio may be processed by Apple, exactly as dictation is.
Listening Mode is optional, and no audio is captured unless the button is pressed. If your setting requires that no speech leave the device, staff can simply not use the button, and every word is still reachable through Topics, Search and Type.
Access methods
Access is by direct touch. Currently, there is no switch scanning, eye-gaze integration, auditory scanning or head-pointer support.