Features
EZConversation is a grid-based AAC app for the iPad, with two key features that most grid-based AACs lack: it listens to the conversation partner, and it builds the grammar for you.
Listening Mode
A wide Tap to Listen button sits along the bottom of the screen. When pressed, the person with whom the user is conversing can say something, and the app will automatically choose likely vocabulary words the user may say.
It also has the ability to detect names and places that are used in the conversation. For example, if one were to ask a user "do you want to go to New York?", EZConversation will automatically add that to the suggested words, even if it is not in the user's vocabulary.
What it does not do is answer, choose words, or speak. It highlights possible sections of vocabulary that a user might say, while still allowing the user to choose any word from the app's or their own added vocabulary, or another word by typing it in.
The Sentence Builder
A short, telegraphic tap sequence comes out as fluent English.
| Tapped | Spoken |
|---|---|
| I · want · cookie | I want a cookie |
| I · want · go | I want to go |
| dog · eat · sandwich | the dog eats a sandwich |
| yesterday · I · go · store | yesterday I went to the store |
| tomorrow · I · eat · pizza | tomorrow I will eat pizza |
| I · no · like · it | I don't like it |
| leg · hurt | my leg hurts |
| dark | it is dark |
| I · headache | I have a headache |
| cat · dog · go | the cat and the dog go |
| I · tired · hungry | I am tired and hungry |
| where · he · go | where does he go |
What it handles (some examples)
- Verb endings — "he go" becomes "he goes".
- Articles — supplies "a", "an" and "the" where English wants one, and leaves them out where it doesn't.
- Missing "is / am / are" — "I happy" becomes "I am happy".
- Negation — a "no" tile before an action produces "don't", "doesn't", "isn't" or "not", as the sentence requires.
- Tense — a time word anywhere puts the whole sentence in the past or future, including "five minutes ago" and "last night".
- Questions — "why I sad" becomes "why am I sad".
- Body parts — "stomach hurts" is always "my stomach hurts", never "the stomach".
- Coordination — two things tapped in a row are joined with "and", since there's no "and" tile to tap.
No function-word tiles, on purpose
There is no my, your or his tile; no and, but or because; no is. Possession is tapped as the plain person, twice: I · like · I · car is "I like my car", and he · like · he · car is "he likes his car".
That is the whole point. The I / me / my distinction is the morphology that the intended audience of EZConversation may struggle with most, which is why the app does not ask the user to learn it.
Where the app inserts a word, it shows as a small grey chip in the message bar, and it can be switched off with a tap.
Finding words
- Instant Words — yes, no, wait, help, I don't know, stop sit in a fixed strip under the message bar and speak the moment they're tapped, without waiting for a sentence. They never move.
- Pinned words — the core vocabulary, on a tab of its own that is one tap away on every screen and always laid out the same way. People, saved phrases and your own words get Home tabs of their own. Fully editable.
- Topics — twenty-five picture-first topic cards, from Food and Feelings to Not Well, School and Money, ordered most generally useful first. The last few topics opened stay as tabs, each with its picture, so moving between them is one tap.
- Search — every word in every topic, plus your own people and custom words.
- Type — an on-screen keyboard with predictions that learn from what this user actually says.
Two keyboard modes
The keyboard commits either Words — each word becomes its own tile and gets the full grammar treatment — or Phrase, where the whole line becomes a single tile the grammar rules leave completely alone.
Phrase mode is the escape hatch for someone typing a finished sentence: the Sentence Builder is designed for telegraphic input, so it would otherwise "correct" a sentence that was already right. A committed phrase behaves like any other tile and mixes freely with tapped ones.
Making it yours
- People & Pets — family, friends, teachers and pets, each with a photo. The photo appears on the tile; only the name is ever spoken. This is usually the highest-value setup step there is.
- Your own words — anything the app doesn't have, with a photo from the camera roll. Add a word straight into a topic, so it turns up where it belongs, and answer a plain-language question about what kind of word it is so the grammar handles it correctly.
- Pinned words — add and remove the always-visible core, with a restore-defaults option. Reorder them by dragging on the grid.
- Saved phrases — whole sentences kept on the home screen, best captured by pressing Save on something that has just been said.
- Reorder — drag tiles into the arrangement that suits this user, on the real grid rather than in a proxy list.
- Recent — the last 50 sentences, each of which can be re-spoken, or loaded back into the message bar as tiles for editing.
The voice
Rather than being chosen from a list of prebuilt voices, the voice is built from a few traits — an age, a vocal quality, and speed, pitch and volume sliders. Every change will play a sample immediately, because the only way to judge a voice is to hear it.
The vocal qualities are labelled Type A, B and C, not by gender. The voice syncs with everything else, so one built on one iPad follows the account to the next.
Access and accommodations
- Tile size — small, medium or large, trading words per screen against target size. Tiles off screen can be scrolled to.
- Hold to preview — press and hold any tile to hear that word without adding it to the sentence. The hold time is adjustable.
- Repeat-tap filtering — ignores a second tap on the same tile within an adjustable window, for tremor or a bouncy touch.
- Speak each word when tapped — optional per-selection audio confirmation.
Access is by direct touch. Currently, there is no switch scanning, eye-gaze integration, auditory scanning or head-pointer support.
Supporter modeling
The best way to teach AAC is to use it yourself in front of the person learning it. But the app is keeping track of what its user says — both to improve word predictions and for progress reports — and a supporter's sentences aren't theirs.
Supporter modeling suppresses both while it's on, with an unmissable orange banner so it's never left on by accident. It resets off every time the app launches.
Backup and reporting
Where a school or clinic has set up an account, the app signs in at launch and quietly backs up people, custom words, pinned words, phrases, recent sentences and settings.
A user can then link a parent or SLP to their account with an invite code, which gives that person a usage report in the browser: how many sentences over time, which topics they came from, vocabulary breadth, and average sentence length by week.
Access is always granted by the user and revocable by them at any time. There is no way to look someone up from the other side.
Symbols
Tiles use OpenMoji for concrete nouns and Mulberry Symbols for abstract AAC vocabulary, both under CC BY-SA, with attribution shipped in the app. Tile colour marks where a word comes from: light grey for the pinned core, dark grey for the current topic, blue for names.