Product names, customer names, internal abbreviations, the way your company spells its own terms: those are the words a generic speech model gets wrong. A shared dictionary teaches them once for everyone, and shared snippets do the same for text people retype all day.
This article covers who can change what, and how the shared set combines with each person's own entries.
Where the shared dictionary lives
Every dictionary entry and every snippet is either personal or shared with the team. Both surfaces show that split:
- Enterprise console — enterprise.sprecho.ai → Dictionary or Snippets, with tabs All, Personal and Shared with team.
- Desktop app — the Dictionary and Snippets screens, with the same scopes.
To share an existing entry, switch it to shared. The entry moves scope: it leaves your personal list and appears in the team list, and both changes reach everyone's devices together.
Who can edit what
| Create shared entries | Edit or delete a shared entry | |
|---|---|---|
| Owner | Yes | Any entry |
| Admin | Yes | Any entry |
| Member | Yes | Only entries they created themselves |
That model keeps a shared list open to contributions from the people who actually notice the missing words, without letting one member silently rewrite what someone else added.
Sharing requires a team. Trying to mark an entry as shared without one is rejected outright rather than quietly saved as a personal entry.
How shared and personal entries combine
Both sets apply at once. There is no override rule, and nothing is hidden:
- Your app reads the union of your personal entries and your team's shared entries.
- The same word can exist once personally and once team-wide. Duplicates are blocked only within one scope, so a personal entry with the same spelling as a team entry is accepted.
- When Sprecho hands the vocabulary to the recognition engine it merges both lists, removes case-insensitive duplicates keeping the first it sees, puts starred entries first, then the most recently updated, and caps the list at 200 words.
A common expectation is that a personal entry overrides the team's version of the same word. It does not — both are in the list, and the first match wins. If a team entry is wrong, fix it in the shared list rather than adding a personal one on top.
Automatic learning respects this too: when Sprecho notices you correcting the same word repeatedly, and that word already exists personally or in the team list in any capitalisation, it refreshes the existing entry instead of creating a near-duplicate.
Shared snippets
Snippets follow the same scope model. A trigger phrase you speak expands into a longer block of text — a standard disclaimer, an address, a closing formula.
- A trigger has to be unique within its scope. The same trigger can exist once personally and once team-wide.
- When several snippets could match, the longest trigger wins, so a specific team snippet is not shadowed by a shorter personal one.
- Formatted snippets — with bold, lists or links preserved on paste — are an Enterprise feature.
Practical advice for a rollout
- Start with the words your recognition actually gets wrong. Ask three or four people for their top ten; the overlap is your first shared list.
- Put company-wide terms in the shared list and leave personal names, client nicknames and role-specific jargon personal.
- Star the entries that matter most. With the 200-word cap, starring is what guarantees a term stays in the list for a heavy dictionary user.
- Review the shared list when products get renamed. A stale entry actively makes recognition worse.