Skip to content

Real-time Translation

Version badge

As of Cognigy.AI 4.10.0 real-time translation has been added to our Conversational AI product.

Real-time translation is a type of translation where not everything has to be pre-translated or pre-localized (see Localization, even if it is done with machine translation, but where the translation machine translates the conversation as it happens.

Virtual Agents are deployed in Cognigy.AI with so-called Endpoints. Webchat, for example, is such an Endpoint and is linked to the currently selected Flow of the Virtual Agent. Real-time translation works on chats and for all our Endpoint types (for example the "Voice Gateway" Endpoint). The Voice Gateway (VG) can be configured in the settings for real-time translation in the same way as for a Webchat.

  • You do not need to have agents that speak all the required languages of your business. You can concentrate on two or three core languages but using Cognigy real-time translation capabilities you can serve customers in a hundred languages.

  • Real-time translation by Cognigy.AI works with any Live Agent system like Genesys, LivePerson, Ring Central,...

  • Cognigy.AI can act as a real-time translation service provider for your global service communications and also work with third-party contact center systems.

We can consider two use cases regarding the real-time translation:

  • Translation between User and Virtual Agent in real-time, even with auto-detection (working for Virtual Agents conversations and also for phone calls).

  • Translation between User and a Human Agent (Live Agent) in real-time, whilst keeping a record of both for compliance purposes.

Real-Time Translation between User and Virtual Agent

In Cognigy.AI Virtual Agents are deployed with so-called Endpoints, for example, the Webchat Endpoint. It may be necessary for customers to make this chat available in another language.

In that case, we have to activate the Real Time Translation Settings.

Endpoint Editor: "Real Time Translation Settings"

  • Go to the Endpoint editor and activate Real Time Translation in the Real Time Translation Settings.

  • Enable Translation using the slider.

  • Select the User Input Language using the drop-down menu.
  • Select Flow Language using the drop-down menu.

  • Save the settings.

  • With a click on "Open Webchat" you can chat with the Virtual Agent.

Example:

Webchat conversation with the preselected user language.

In Webchat, we're assuming that the language that is being typed in is Russian. The Virtual Agent content is being translated into Russian in real-time.

Webchat Real-time translation and preselected language for user inputs

The whole conversation of everything is translated including e.g. calendar of the date picker (day, month, week,...).

Real-Time Translation Settings

The configuration options and their functions are listed below:

Parameter Description
Configured Translation Provider The currently selected translation provider.
Enable Translation Whether Translation is enabled for the Endpoint.
User Input Language The language of the user inputs.
Flow Language The language of the Flow outputs.
Set user input language on execution count If the input language is set to 'Auto-Detect', it will be fixed to its current value on this execution count.
No-Translation Marker Do not translate texts encapsulated in this marker (e.g. 'I love the Cognigy.AI platform' will not translate 'Cognigy.AI').
Always remove No-Translation Markers Whether we remove No-Translation Markers, even if the translation is not enabled.
Prevent payloads from being translated If active, prevents all user inputs based on payloads to be translated.

Auto-Detect User Inputs

We can not just have a preselected language in the settings, but we can also say Auto-Detect(auto) as User Input Language.

Auto-detected user input (german language)

When auto-detection for user inputs is activated, then we have an extra setting: Set user input language on execution count.

  • Use the slider to set the option Set user input language on execution count as you want. The slide bar can be set between "0" and "5".

Example: slider is set to "1" which means: Whatever language the user is using on their first input is the language that the Virtual Agent will use from then on. These settings work with every language we support.

Endpoint Editor: "Auto-Detect" setting

Real-Time Translation between User and Live Agent

  • In the Endpoint settings set "Flow" to "Handover", which will hand the conversation over to a Live Agent (human agent).
Webchat Editor: set "Flow" to "Handover"

Example:

  • User language is assumed in the German language.
  • Handover Flow and Agent use the English language.

The Webchat conversation of a user with the Virtual Agent starts. At any point of the conversation the user would like to talk to a human agent like "I would like to talk to a human Agent" and Cognigy Virtual Agent reacts to the Intent, stops the Flow execution, and leads the conversation over to a Live Agent who can overtake and continue with the conversation.

Webchat conversation with a users handover request

Chat flow process between the user and Live Agent using real-time translation is as follows:

  • The Live Agent will see in his chat the real-time translated user inputs in English and replies in English also.

  • The user will get the real-time translated Live Agent replies in German in his Webchat application and replies in German.

Live Agent (English) conversation with a user (German) in real-time translation

With a click on Resolve, the Live Agent can close the conversation and the user is back to the Virtual Agent conversation as it was before.