Skip to contentSkip to navigationSkip to topbar
On this page

Language Operators


(information)

Generative Custom Operators now available in public beta

Use LLM-powered Language Operators for sophisticated and flexible natural language understanding tasks. For more information, see the Generative Custom Operators docs.

Language Operators use artificial intelligence and machine learning technologies to provide additional analysis and insights on your Transcripts.


Operator types

operator-types page anchor

You can use two Language Operator Types with a Voice Intelligence Service:

  • Pre-built Language Operators: Twilio creates these in-house or trains them using third-party AI models. Twilio trains Pre-built Operators across a wide swath of data, mapping them to information that's independent of your industry or use case. These are Twilio-maintained Operators that you can't customize.
  • Custom Operators: You can also create customized Operators. They're Operators specific to your use case and data. We offer two types of Custom Operators:
    • Generative: Use LLM-powered analysis for sophisticated and flexible natural language understanding tasks.
    • Phrase Matching: Specify a keyword or a phrase to find in a Transcript or use as a category for classifying your Transcripts.

Operator actions overview

operator-actions-overview page anchor

Operators perform specific actions on a conversation or a sentence within a conversation. There are four types of actions that an Operator can perform.

ActionDescriptionExample
GenerativeUse an LLM to generate text or JSON for sophisticated and flexible natural language understanding tasksScore a call based on multiple custom-defined categories of agent performance.
ClassifyClassify a conversation into a predefined categoryClassify if the call transferred to another agent.
Phrase matchingDetermine if an event occurred or if someone mentioned a piece of data or a phrase during a conversation.Determine whether an agent informed a customer that they're recording the call.
Classify-ExtractExtract relevant information like entities from a conversation and classify those entities into a predefined categoryIdentify and extract unique entities such as locations, organizations, consumer goods or people, quantities, and times.

Custom Operators are available for all Transcription languages.

Pre-built Operator language support varies depending on the Transcription language and the Operator. For detailed information, refer to the Pre-built Operator table.