Language Profiles

Customer experience functionalities such as live translation, transcription and others rely on the system's ability to handle multiple languages in a business workflow.

Language specific services involve two participant roles in a conversation: a customer and an agent, potentially being fluent in different languages. We offer dedicated language profiles for each, therefore a Language Profile designates a language for the customer or the agent as well as translation service providers (STT/TTS, voice models, etc). It also specifies what use cases or language services the profile can be used for (e.g. voice or chat translation) as well as business specific key phrase vocabularies and the ability to correct translation errors by text manipulation.

Create agent language profile

Navigate to AI Administration > Language Profiles on the portal.
Click the Agent tab and Create an agent language profile.
Enter the Language profile name, select a Language from the list, then click Next.
Select the feature you would like this profile to be used for. Some features might need further details to be provided, others might not. You may enable multiple features, however, the list of available providers and settings will be trimmed down to ensure compatibility, as some providers are not available for some features.
Click Create. The new profile is saved and its details show up next to the list.

Agent-side Language profile features

The following features can be configured for an agent language profile:

Call transcription

This transcribes voice calls into text in realtime. You must select one of the available providers for the chosen language and, optionally, you may select Text manipulation rules and a vocabulary. This feature is paired with the next best action feature for agents, see the Agent Assist for more details.

Note

Read more about using call transcription with Next Best action in the agent guide here.

Assign the agent language profile to your agents. See the agents section for more details on how to set the Spoken language and the language profile.

Call Translation

Enables the agent side for call translation and defines the configuration of the agent. You must select a speech-to-text and a text-to-speach providers available for the chosen langauge. See translation language support for additional information.

You must also select a Text To Speech voice model, this is the voice the customer will hear when the call is translated.

You may also configure additional parameters such as a vocabulary and text manipulation rules.

Moreover, for the text-to-speech side, you may opt-in to control the digit reading speed and other settings, depending on the provider and voice model you selected.

Please note that the Call Translation feature is available for inbound and outbound calls, but not available for campaign-related calls, transfers and conferences. See the agent guide for inbound and outbound here.

Translation pairs are created automatically between agent language profiles and the customer language profile, you may customize the translation provider by selecting an agent profile and navigating to the Translation Engine tab. Here is where you can customize the select agent profile's translation pairs with any other customer language pair. In some cases, other features may be available depending on the provider, such as Dictionary, change the formality and more.

Assign the agent language profile to your agents. See the agents section for more details on how to set the Spoken language and the language profile.

Chat Translation

Enables the agent side for chat translation. You do not have to select any other details, but to turn on chat translation for a given agent, this feature needs to be turned on in the language profile.

Assign the agent language profile to your agents. See the agents section for more details on how to set the Spoken language and the language profile.

Read more about the agent experience for chat translation here.

Email Translation

Enables the agent side for email translation. You do not have to select any other details, but to turn on email translation for a given agent, this feature needs to be turned on in the language profile.

Assign the agent language profile to your agents. See the agents section for more details on how to set the Spoken language and the language profile.

See the agent guide for details on the agent experience.

Create customer language profile

Navigate to AI Administration > Language Profiles on the portal.
Click the Customer tab and Create a customer language profile.
Enter the Language profile name, select a Language from the list, then click Next.
Select the feature you would like this profile to be used for. Some features might need further details to be provided, others might not. You may enable multiple features, however, the list of available providers and settings will be trimmed down to ensure compatibility, as some providers are not available for some features.
Click Create. The new profile is saved and its details show up next to the list.

Features available for the customer language profile

The following features can be configured for a customer language profile

Call translation

Enables the customer side for call translation and defines the configuration of the customer. You must select a speech-to-text and a text-to-speach providers available for the chosen langauge. See translation language support for additional information.

You must also select a Text To Speech voice model, this is the voice the agent will hear when the call is translated.

You may also configure additional parameters such as a vocabulary and text manipulation rules.

Moreover, for the text-to-speech side, you may opt-in to control the digit reading speed and other settings, depending on the provider and voice model you selected.

Assign the customer language profile to for your customers. Depending on the direction of the call.

For outbound calls, this is set in the Agent Workspaces' voice tab where you may configure additional translation settings.

For inbound calls, this is set in the voice workflow's set customer language node, depending on which path you wish to enable translation.

Chat translation

Enables the customer side for chat translation. Set the language profile in your chat workflow's set customer language node.

Email translation

Enables the customer side fro email translation. Set the language profile in your email workflow's set customer language node

Recording Transcription

This enables the transcription of customer calls. You must select an Offline transcription provider. Then, you may set this language profile in the tenant's settings to apply it globally or in a voice workflow's set customer language node to override the global configuration.

Please note that the customer perspective of the conversation must be recorded, otherwise, the Recording transcription will not produce any results, in cases where the recording is disabled or set to only record the agent perspective. For clarity, the agent perspective does include the customer speech, but is limited to the task handled by the agent, not showing the full perspective of the customer and not including workflow messages, other agents, conversations with AI bots and so on.

Voice Bot

This enables the configuration for voice conversational bots. You must select a Speech to Text, Text to Speech provider and a voice model from the available list for the chosen language.

You may also configure additional parameters such as a vocabulary and text manipulation rules.

Moreover, for the text-to-speech side, you may opt-in to control the digit reading speed and other settings, depending on the provider and voice model you selected.

Then, you must select this language profile in your voice workflow's set customer language node, on the path that reaches a conversational bot node. If multiple set customer language node exit on the same path pointing to a conversational bot node, then the language profile configuration of the last set customer language node will be used.

Workflow TTS

This feature enables the configuration of announcements Text To Speech provider and voice model. You must use this configuration in a set customer language node in your voice workflows.

If multiple paths, each having a set customer language node with the Workflow TTS enabled reach the same announcement(s), then the administrator can configure per announcement a message for each language profile.

You may opt-in to control the digit reading speed and other settings, depending on the provider and voice model you selected.

Delete a language profile

Navigate to AI Administration > Language Profiles on the portal.
Click the tab for the designated language profile type (Agent or Customer).
Find and click the language profile you would like to remove.
Click Delete at the top, and confirm your intent.

Note

The item will be removed even if it is being referenced by other configurations. Make sure it is not in use by an important business entity.

Create key phrase vocabularies

Modern translation services rely on AI models for extremely fast and high quality results. However, being reliant on generic text and speech data, the result could contain less precise or inadequate phrases for the particular business context. The vocabulary of key phrases aids the language services in using precise phrases.

While at the Language Profiles menu, click the Vocabulary tab to see their list.
Click Create and type a name for the new vocabulary.
Type a phrase in the Add a new phrase field and click Add to insert phrases individually.
Click Import to upload a list of phrases from a CSV or text file in one operation and decide whether duplicates found will be replaced or appended.
Click Create to create the new vocabulary.
Select an Agent or Customer profile, and select from the Speech To Text Provider Vocabulary dropdown.

Delete a key phrase vocabulary

Navigate to AI Administration > Language Profiles on the portal and click Vocabulary.
Click the desired item and click Delete on the details panel.
Confirm your intent.

Create text manipulations

For translation services involving STT/TTS workflows the system offers the ability to address typical mistranslations of business specific identifiers and other tokens using powerful regex statements, including capture groups.

For example, if the input string is Inc 1234 but the expected outcome is INC1234 (a business token, e.g. an ID), then your Regex field should be: Inc\s(\d{4}) while the Replace field should be INC$1.

Such manipulations are executed on the chunk of text the STT engine generated from the last utterance of the relevant party, and before the text is translated to the target language and handed off to the TTS engine.

While at the Language Profiles menu, click the Text Manipulations tab to see the list.
Click Create and type a name for the new text manipulation ruleset.
Type a Regex to match a specific word or phrase in the text generated by the STT provider. Also type a Replacement word or phrase and an optional Description for this rule.

When typing the regex and replacements, mind the language this text manipulation ruleset is expected to be used with, and apply language specific writing notations when relevant. Do not use accented characters as those are not necessarily available or translated precisely by the translation services. As usual with AI enforced features, proper testing and experimentation is due before you roll any config into production.
Click Add to insert the rule into the ruleset list.
Repeat these steps to add all rules to the ruleset and then click Create to save the new text manipulation ruleset.
Edit or create the corresponding agent and/or customer language profile and select this ruleset in the Text Manipulation List dropdown, then create or save the profile.

Delete text manipulations

While at the Language Profiles menu, click the Text Manipulations tab to see the list.
Click the desired item and click Delete on the details panel.
Confirm your intent.

Digit reading options (TTS)

Some providers support passing parameters related to the reading speed of the digits, this is particularly useful for situations where a number would have to be read digit by digit instead of the entire number at once, such as reading out loud confirmation numbers, order IDs and so on.

When Read digit by digit feature is turned on, the reading speed can be controlled, these are arbitrary values sent to the TTS provider.

This feature is not available for all providers.

Additional provider settings (TTS)

Some providers support customizing the Text-To-Speech experience in more detail than just the reading speed of digits, these settings are:

Stability: This determines how stable and consistent the voice is, and the randomness between each generation. Lower values tend to introduce broader emotional range for the voice, higher value results in monotounous voice with limited emotion. Values range from 0 to 1 with double decimals. 0 represents the lowest value and should not be confused with disabled.
Use speaker boost: This setting, when turned on, will prompt the TTS provider to boost the similarity to the original speaker at the cost of latency as it requires more computing power. Turn this on only if increased latency is acceptable for your scenario.
Similarity boost: Turn this setting on and set a value between 0 and 1 with 2 decimals to determine how close should the AI adhere to the original voice when attempting to replicate it. Pair this with use speaker boost.
Style: Turning this on and setting a value higher than 0 will increase the latency but will prompt the provider to attempt to match the speaking style of the original speaker.
Auto-detect language: The provider will auto-detect the langauge and in some cases, it may produce better results.

Please note that these settings are fully under the control of the provider, the product allows setting them from the administration portal and results may differ from language to language.

Language profile features checklist

To make sure you got everything you need for a feature configured here to work, consult the checklists.

Call translation

[ ] You must have an agent language profile created with the call translation feature turned on.
[ ] Your agent language profile must be assigned to your agents alongside their spoken languages. If their spoken language does not match your language profile language you will not be able to select it.
[ ] You must have a customer language profile created with the call translation feature turned on.
[ ] You must assign the customer language profile in the agent workspace's voice tab, if you wish to enable translation for outbound calls.
[ ] You must use a set customer language node in your voice workflow that has customer language profile containing the call translation feature on the path that reaches a queue.
[ ] The language of the customer must not be one of the spoken languages assigned to the agent, in that case, translation is disabled.

Chat Translation

[ ] You must have an agent language profile created with the chat translation feature turned on.
[ ] Your agent language profile must be assigned to your agents alongside their spoken languages. If their spoken language does not match your language profile language you will not be able to select it.
[ ] You must have a customer language profile created with the chat translation feature turned on.
[ ] You must use a set customer language node in your chat workflow that has customer language profile containing the chat translation feature on the path that reaches a queue.
[ ] The language of the customer must not be one of the spoken languages assigned to the agent, in that case, translation is disabled.
[ ] The chat conversation is not an outbound chat. Chat translation is not supported for outbound chats.

Email Translation

[ ] You must have an agent language profile created with the email translation feature turned on.
[ ] Your agent language profile must be assigned to your agents alongside their spoken languages. If their spoken language does not match your language profile language you will not be able to select it.
[ ] You must have a customer language profile created with the email translation feature turned on.
[ ] You must use a set customer language node in your email workflow that has customer language profile containing the email translation feature on the path that reaches a queue.
[ ] The language of the customer must not be one of the spoken languages assigned to the agent, in that case, translation is disabled.
[ ] The email is not an outbound email.

Voice bot

[ ] You have a conversational bot node configured in your voice workflow.
[ ] You have a set customer language node in your voice workflow with the voice bot feature turned on, on a path that leads to your conversational bot.

Recording transcription

[ ] Your business unit configuration is set to record the Customer side or Both.
[ ] You have assigned a customer language profile with the recording transcription feature turned on to your tenant or your voice workflow's set customer language node.
[ ] The customer language matchesthe language of your language profile, otherwise the recording transcription will produce unusable results.
[ ] You wait up to 30 minutes for the recording and transcription to be available.

Workflow TTS

[ ] You must have a valid customer language profile with the Workflow TTS feature turned on.
[ ] You must use a set customer language node in your voice workflow, with the customer language profile containing the Workflow TTS feature selected, on a path pointing to an announcement of any type (bot, auto attendant, etc)
[ ] The announcement must use TTS not WAV and the text for the language profile is set.

Features typically require at least the customer's language profile to be designated in advance. Features involving a human agent will also need their language profiles set up as well for normal operation.

See Language services providers list for further details.