Documentation Index

Fetch the complete documentation index at: https://docs.natterbox.com/llms.txt

Use this file to discover all available pages before exploring further.

Use SSO to log in and access the Help Center, where you can create and manage your support tickets to reduce your resolution times.

AI Advisor Transcription Configuration

Prev Next

Please note

For existing Natterbox AI users, in order to get access to these features you will need to be using the latest version of the Natterbox App. For more information about getting setup please speak to your Natterbox customer success manager.

Language Support

It's now possible to transcribe your AI Advisor calls in multiple languages and dialects, please note it's recommended to use the dialect specific to the region that your calls are being handled  where possible as this will produce higher quality transcriptions and result in better data produced for Insight analytics.

Please make sure you are on at least v1.316 of the Natterbox App in order to access the following languages:

  • Bulgarian

  • Catalan

  • Chinese

    • Chinese (China)

    • Chinese (Taiwan)

    • Chinese (HongKong)

    • Chinese (Simplified)

    • Chinese (Traditional)

  • Czech

  • Danish

  • Dutch

    • Flemish / Dutch (Belgium)

  • English

    • English (Australia)

    • English (India)

    • English (New Zealand)

    • English (UK)

    • English (US)

  • Estonian

  • Finnish

  • French

    • French (Canada)

    • French (France)

  • German

    • German (Switzerland)

  • Greek

  • Hungarian

  • Indonesian

  • Italian

  • Japanese

  • Korean

    • Korean (South Korea)

  • Latvian

  • Lithuanian

  • Malay

  • Norwegian

  • Polish

  • Portuguese

    • Portuguese (Brazil)

  • Romanian

  • Russian

  • Slovak

  • Spanish

    • Spanish (Latin American)

    • Spanish (Spain)

  • Swedish

    • Swedish (Sweden)

  • Thai

    • Thai (Thailand)

  • Turkish

  • Ukrainian

  • Vietnamese

By default all calls will be transcribed into English so you must ensure AI Advisor is configured to use the language your agents and customers will be conversing in.

Multilingual

For calls where it’s unclear from the start which language the conversations will be or for calls where a mixture of languages are used it’s recommend to use the Multilingual option.

The following 10 languages are supported : English, Spanish, French, German, Hindi, Russian, Portuguese, Japanese, Italian, and Dutch.

Please note whilst it’s possible to transcribe and produce AI analysis for the above languages, the following languages are not supported for Category detection :

  • Bulgarian

  • Catalan

  • Czech

  • Estonian

  • Latvian

  • Lithuanian

  • Malay

  • Multilingual (Spanish + English)

  • Polish

  • Slovak

  • Ukrainian

Configuring different languages for AI Advisor

There’s three levels where the transcribed language can be set for Insight, at an account level, user level and routing policy level.

By default the language will always be English, otherwise the hierarchy of how the language setting is overwritten goes in the order : account, user then finally routing policy.

Account Level
  1. Navigate to Account Settings

  2. Under Insight Settings

  3. Set the Transcription Language

User Level
  1. Navigate to Natterbox Users

  2. Edit a user’s record

  3. Under Insight Settings

  4. Set the Transcription Language

Routing Policy Level
  1. Navigate to Routing Policies

  2. Open the call policy (or policies) in which Insight is being used

  3. Edit the Record and Analyse call component (This component is responsible for enabling Insight analysis on all calls that route via that policy)

  4. Click on the Insight tab from within the settings

  5. Ensure the Transcription Engine field has been set to Latest (as only the latest model supports the non-english languages)

  6. Click on the dropdown Analysis Language and choose the appropriate language

  7. Click Save on the policy editor

Note when you configure a Category, unless you’re using the Multiple language option you must ensure it matches the language set within the Record and Analyse component otherwise the Category won’t trigger. Matching the language between the transcription and the Category will reduce the amount of false positives triggered in Insight.


FAQ

Why don't I have access to any non-English languages from the component settings?

If you only have access to English variants, you may need to switch the transcription engine to use the latest model.

From the field Transcription Engine dropdown select Latest, now you should have access to all languages.

In the Search page, why do my transcripts contain lots of nonsensical text?

The spoken language may be different to the selected transcribed language, please ensure that the language set within the Record and Analyse component is set to the matching language.

Are you able to auto-detect the language being spoken and automatically apply the correct language transcription?

Yes, use the multilingual option.

If my call has been transcribed in the incorrect language, can I request Natterbox to re-transcribe it in the correct language?

No, calls must be correctly configured to the correct language before they are processed.

Why are my calls not being transcribed in the selected language?

If you discover your calls are not being transcribed in the selected language as set from within the Record and Analyse component, first you will need to confirm that the affected calls are in fact routing through the call policy that has this component.

If they’re not, you may need to either change the routing so they go via the desired policy or you may need to add a new Record and Analyse component and set the language within the policy that the calls are currently routing via.

Natterbox can help advise on how to best configure this.

If calls are still not transcribed in the selected language then please reach out to Natterbox support.

Why are my calls not transcribing at all?

If you have selected one of the non-english languages, but the transcription engine has been switched back to the Legacy model, it will not be compatible.

You will see from the Record and Analyse component that the language is unset.

You will need to either choose an English language option from the ones provided otherwise it’s recommended to switch to the Latest model and choose a language before saving.

How do I set these languages using a LUA script?

See below for the language mapping:

  • Bulgarian: bg

  • Catalan: ca

  • Chinese: zh

    • Chinese (Simplified): zh-CN

    • Chinese (Hong Kong): zh-HK

    • Chinese (Taiwan): zh-TW

  • Czech: cs

  • Danish: da

    • Danish (Denmark): da-DK

  • Dutch: nl

  • English: en

    • English (US): en-US

  • Estonian: et

  • Finish: fi

  • Flemish: nl-BE

  • French

    • French (France): fr

    • French (Canada): fr-CA

  • German: de

    • German (Switzerland): de-CH

  • Greek: de

  • Hindi: hi

  • Hungarian: hu

  • Indonesian: id

  • Italian: it

  • Japanese: ja

  • Korean: ko

    • Korean (South Korea): ko-KR

  • Latvian: lv

  • Lithuanian: lt

  • Malay: ms

  • Norwegian: no

  • Polish: pl

    • Portuguese (Brazil): pt-BR

  • Romanian: ro

  • Russian: ru

  • Slovak: sk

  • Spanish

    • Spanish (Spain): es

    • Spanish (Latin American): es-419

  • Swedish: sv

    • Swedish (Sweden): sv-SE

  • Thai: th

    • Thai (Thailand): th-TH

  • Turkish: tr

  • Ukrainian: uk

  • Vietnamese: vi

What services support the transcription of calls?

Service Cloud Voice and AI Advisor both support the ability to transcribe calls

Service Cloud Voice dynamically transcribes live calls and Insight transcribes calls as a post call action. Please see our SCV chapter for more info on this.

Where are transcriptions stored?

For AI Advisor transcriptions our stored on the Natterbox platform, in the same way call recordings are.

For Service Cloud Voice all transcriptions are stored within Salesforce.

How long are transcriptions stored for?

AI Advisor transcriptions follow the same archiving policy as call recordings.

For Service Cloud Voice, Natterbox doesn't keep any copies of transcriptions, they're solely stored in Salesforce.

How accurate are the transcriptions?

It's important to first note that the accuracy of a transcription is heavily reliant on many factors surrounding the quality of a conversation, for example :

  • The internet connection of either party

  • The microphone quality of either party

  • The selected language dialect (i.e. The Australian English option will produce better quality transcriptions for Australian calls than the generic English option)

  • The content of the conversation, such that similar sounding words and the use of slang could reduce the overall quality

  • Cross talk

  • Background noise

With those factors in mind we've seen transcription accuracy to average about 80%.

It's worth noting that the AI language models will gradually improve upon over time, so we expect that accuracy value to also increase on average.