Skip to main content

Dictionary management best practices

Dictionary management helps administrators and analysts improve transcription accuracy by adding organization-specific vocabulary, such as brand names, acronyms, and internal terminology.

The dictionary options available to you depend on the transcription engine that your dialect uses. Understanding the difference between Sounds Like and Display As can help you choose the appropriate approach for recognition and formatting issues.

Understand transcription engine differences

Genesys Cloud uses different transcription engines, and the dictionary fields that appear depend on the engine that supports the selected dialect.

FieldGenesys native voice transcriptionExtended Voice Transcription Services (EVTS)
TermYes, up to 30 charactersYes, up to 30 characters
Sounds LikeYes, up to 10 entries per termNot available
Display AsYes, up to 50 charactersYes, up to 50 characters
Boost valueAPI onlyNot available
Note: Verify the current availability of Display As for Genesys native voice transcription before publishing the article. The source document contains conflicting information about its availability.

Dictionary entries are specific to the transcription engine and dialect. If you move a dialect from one transcription engine to another, dictionary entries are not automatically migrated. Recreate the applicable entries for the new engine.

Because field availability can change, confirm the options for your dialect in Admin > Quality > Dictionary Management > Add Term before creating your dictionary strategy.

Understand the Sounds Like and Display As fields

The Sounds Like and Display As fields address different transcription issues.

Sounds LikeDisplay As
PurposeHelps the engine recognize a termControls how a recognized term appears in the transcript
Affects recognitionYesNo
Affects transcript appearanceNoYes
Typical use caseA term is missed or misheardA term is recognized but displayed incorrectly

Sounds Like

Use Sounds Like when Genesys native voice transcription does not consistently recognize a term or recognizes it incorrectly.

Sounds Like provides phonetic alternatives that help the transcription engine identify how a term is pronounced. It does not control how the recognized term appears in the transcript.

For example:

  • Term: Neurological
  • Sounds Like: neuro logical
  • Sounds Like: euro logical

Use Sounds Like when a term is:

  • Missed in transcripts.
  • Split incorrectly.
  • Consistently misheard.

Display As

Use Display As when the transcription engine recognizes a term, but the term does not appear in the desired format.

Display As changes how the recognized term appears in the transcript. It does not affect whether the transcription engine recognizes the term.

For example:

TermDisplay As
four o one k401K
adpADP

Display As is useful for correcting:

  • Capitalization.
  • Spacing.
  • Punctuation.
  • Numeric formatting.

Use Display As to correct consistent transcription errors

In addition to formatting recognized terms, you can use Display As to correct a consistent, repeatable transcription error.

If the transcription engine repeatedly produces the same incorrect term or phrase, enter the incorrect transcription as the term and the desired text as Display As. When the engine recognizes that exact term, Genesys Cloud displays the corrected value instead.

For example, assume that your product is named Vantage Health, but transcripts consistently show Vintage Health.

Configure the dictionary entry as follows:

FieldValue
TermVintage Health
Display AsVantage Health

When the transcription engine produces Vintage Health, the transcript displays Vantage Health instead.

Important: Use this approach only for consistent errors that you observe in actual transcripts. Display As matches the specific Term value; it is not a general substitution rule.

If the transcription error varies, create a separate entry for each observed variation. For example, if the same term appears incorrectly as both Vintage Health and Vintage Hell, each variation requires its own entry.

Follow dictionary formatting requirements

Formatting requirements differ between Genesys native voice transcription and EVTS.

Genesys native voice transcription

For English:

  • Term and Sounds Like support letters, apostrophes, and spaces.
  • Digits, accented characters, and other punctuation are not supported.
  • A term can contain up to 30 characters.
  • You can add up to 10 Sounds Like entries for each term.
  • Matching is case-insensitive.

Other languages have their own supported character sets.

Extended Voice Transcription Services

For EVTS:

  • A term can contain up to 30 characters.
  • Do not use digits in a term. Spell the numbers out instead. For example, use zero two instead of 02.
  • Do not use unsupported special characters or hyphens in Term.
  • For acronyms, add a period after each letter, including the final letter. For example, use F.B.I..
  • Enter multi-word terms with spaces between words.
  • Display As can contain up to 50 characters and supports digits and punctuation.

For a brand name that contains formatting that Term does not support, enter an allowed version in the Term field and use Display As for the desired transcript formatting.

Allow time for dictionary updates

Dictionary changes typically take effect within approximately 15 minutes. Changes apply only to new interactions. Conversations that are already in progress do not receive dictionary changes retroactively.

After you add or update an entry, check its status rather than assuming that the change is immediately active.

Possible statuses include:

  • Pending
  • Active
  • Failed
  • Invalid

Use the following workflow when you create or maintain dictionary entries:

  1. Review actual transcripts and identify terms that require improvement.
  2. Determine whether the issue is a recognition problem or a display problem.
  3. For recognition problems, use Sounds Like when available.
  4. For formatting or repeatable correction problems, use Display As when available.
  5. Add the entry in Admin > Quality > Dictionary Management and select the appropriate transcription engine and dialect.
  6. Save the entry and check its status.
  7. Allow up to 15 minutes for the change to become active.
  8. Review new transcripts to confirm that the dictionary entry produces the expected result.

Avoid common dictionary management issues

When you manage dictionary entries:

  • Do not treat Sounds Like and Display As as interchangeable. Sounds Like affects recognition; Display As affects transcript appearance.
  • Do not enter unsupported numeric or hyphenated terms when you can use an allowed Term and the desired formatting in Display As.
  • Do not assume that dictionary changes take effect immediately or update interactions already in progress.
  • Do not create Display As correction entries based on guesses. Create them from transcription errors that you observe consistently.
  • Remember that entries are specific to the transcription engine and dialect.
  • Check the entry status after you save it to make sure that the entry becomes active.

Maintain your dictionary

Review your dictionary regularly as your organization’s terminology changes.

Consider reviewing dictionary entries when:

  • You introduce new products, services, or campaigns.
  • QA or analytics reviews identify recurring transcription errors.
  • Existing terminology is no longer relevant.
  • A dialect moves to a different transcription engine.

A quarterly review can help keep the dictionary relevant and manageable. Remove entries that are no longer needed because unused entries still count toward per-dialect capacity and increase administrative overhead.