Conversation Analyzer

Provisioned feature

The feature described on this page may not be enabled for all accounts. Access depends on your current license level or specific account configuration.

1 Overview
- 1.1 Language and profile
- 1.2 Components
- 1.3 AI-powered interaction summaries
2 Limitations
3 Prerequisites
4 Language determination logic
- 4.1 Supported language formats
5 Profile determination logic
6 Data source logic
- 6.1 Priority hierarchy
- 6.2 Inbound line and outbound interactions
7 Integrated insights

Overview

Conversation Analyzer (CA) processes voice and digital interactions to provide detailed, automated analysis of customer-agent interactions, including:

Voice: Calls and voicemails. For more information, see Analyzing voice interactions and voicemail recordings.
Digital: Email, SMS, WhatsApp, and webchat. For more information, see Analyzing digital interactions (webchat, SMS, WhatsApp, and email).

Language and profile

For CA to analyze an interaction, your interaction plan must be configured with two key settings: a language and a profile. The language determines which transcription engine CA uses to process the interaction, while the profile defines the set of category rules applied to the transcript to identify topics of interest - such as compliance statements, complaints, or competitor mentions. If either setting is missing at the end of the interaction, CA does not perform the analysis.

Components

Once both profile and language are configured, CA automatically processes the interaction when it ends, producing the following components:

Intelligent categorization: Categories and subcategories are identified based on the profiles defined in the Category Editor. The UI automatically sorts identified categories to the top for faster review.
Transcript & highlights: A full text record of the interaction with category-specific keyword highlighting.
Visual timelines:
- Voice: Shows agent talk, customer talk, silence, and talkover.
- Digital: A chronological timeline of message events and detected insights.
Sentiment: Includes sentiment progression graphs for eligible interactions.
For more information, see Sentiment analysis in voice and digital interactions.

AI Insights: Includes AI-generated summaries (including key topics and follow-up actions) for eligible interactions.

Analyzed interactions are marked with a document icon in the Type column of the Interaction Content area, which contains all the interactions for your account. For more information, see Interaction Content.

AI-powered interaction summaries

Interaction eligibility

AI summaries are eligible for interactions where:
- The conversation time is at least 10 seconds long (voice interactions).
- There is at least one word in the transcript (digital interactions).
- Language for the interaction is properly configured. See Language determination logic for more information.

AI-powered interaction summaries are available for both voice and digital interactions (see Analyzing digital interactions (webchat, SMS, WhatsApp, and email) for more information). Summaries are generated in the language configured for the interaction.

Limitations

Parameter	Voice	WhatsApp, chats, and webchats	Email
Duration Limit	First 120 minutes
Content Scanned	Audio stream	Message text	Subject & Body text only
Attachments analyzed?	N/A

Prerequisites

Conversation Analyzer must be enabled for your account.
Analyzing digital interactions must be enabled for your account.
Agents must have the following switches enabled in their User Admin settings:
- Conversation Analyzer for chats: Set to Yes for chat and email interactions to be analyzed.
- Transcribe calls: Set to Yes for voice interactions to be analyzed.
The Call Recording applet must be placed before the TakeMessage or Voicemail applet in your interaction plan for voice.

Language determination logic

Conversation Analyzer cannot automatically detect the language of an interaction. To ensure accurate transcription and analysis, every interaction must be tagged with a supported language code. The system uses the value stored in the Language data source at the end of the interaction to select the appropriate transcription/summarization profile.
See the Set Data Source applet section to learn how to configure the language for your interaction plan.

Language limitations

Conversation Analyzer processes interactions in one language only. If an interaction contains multiple languages, only the fragments matching the tagged language will be accurately transcribed/summarized.
If the Language field is empty:
- Analysis is not performed.
- Summarization is not provided; the system displays an error.

Supported language formats

The Language data source must contain a supported code in the specific ISO format (e.g., ll-cc).

Supported language Required format Arabic ar-sa Bashkir ba-ru Basque eu-es Belarusian be-by Bulgarian bg-bg Catalan ca-es Croatian hr-hr Czech cs-cz Danish da-dk Dutch nl-nl English (Australia) en-au English (United Kingdom) en-gb English (United States) en-us Estonian et-ee Finnish fi-fi French fr-fr Galician gl-es German de-de Greek el-gr Hindi hi-in Hungarian hu-hu Indonesian id-id Italian it-it Japanese ja-jp Korean ko-kr Latvian lv-lv Lithuanian lt-lt Malay ms-my Marathi mr-in Mongolian mn-mn Norwegian no-no Polish pl-pl Portugese pt-pt Romanian ro-ro Russian ru-ru Slovakian sk-sk Slovenian sl-si Spanish es-es Swedish sv-se Tamil ta-in Thai th-th Turkish tr-tr Uyghur ug-cn Ukrainian uk-ua Vietnamese vi-vn Welsh cy-gb

Transcription accuracy

Always ensure that your interaction plan correctly captures the language before the interaction ends.
If your plan allows for language switching (e.g., "Press 2 for Spanish"), a Set Data Source applet must update the Language tag to es-es to ensure Conversation Analyzer uses the correct transcription engine.

Profile determination logic

Conversation Analyzer cannot automatically detect which category profile to apply to an interaction. To ensure the correct business rules and keywords are used, every interaction must be tagged with a profile name. The system identifies this using the value stored in the ConversationAnalyzerProfile data source at the end of the interaction. For more information, see Category Editor for Conversation Analyzer.

Profile data requirements

To be recognized by the system, the profile name must meet the following technical criteria:

Format: Must be a string.
Length: No longer than 32 characters.
Allowed Characters: Numbers, letters, hyphens (-), and underscores (_).

Troubleshooting

If an interaction is showing "undefined" for the profile in the player metadata, check the interaction plan.

If you are using a Set Data Source applet, ensure that:

The variable name is exactly ConversationAnalyzerProfile (case-sensitive).
The Set Data Source applet is triggered before the interaction ends.

Data source logic

VCC determines the final values for both Language and ConversationAnalyzerProfile at the exact moment an interaction ends. Because these tags are dynamic, they can be changed as a conversation progresses.

This hierarchy applies to both voice and digital interactions, and covers:

Account defaults.
Inbound line or callback number settings.
Interaction plan applets (Set Data Source or Data Connector).

Settings configured by Vonage

Certain foundational settings can only be configured by Vonage:

Setting the Account Default Language or Categorization Profile.
Associating specific Languages or Profiles with inbound lines or callback numbers.

Priority hierarchy

Tag override hierarchy

This priority hierarchy applies identically to both the Language and ConversationAnalyzerProfile data sources.

Priority	Logic step	Behavior
1 (highest)	Interaction plan applets	Overrides all others. Values set via Set Data Source or Data Connector are the final tags used for analysis.
2	Inbound line or callback number	Overrides the Account default. Tags are determined by the specific phone number or digital entry point used.
3 (lowest)	Account default	This is the initial value at the start of an interaction if no other overrides exist.
Special case	Re-routed interactions	If an interaction is routed to another number or a different interaction plan: It carries the current value of the tags with it. The language or profile of the new line does not override the existing values. However, any applets in the new interaction plan (Priority 1) can still override the tags.

Example
The Global Support scenario

The following table illustrates how a tag evolves as a call moves through the system. In this example, a French-speaking customer calls a company whose default language is English.

Interaction stage	Component	Setting applied	Resulting tag	Description

Interaction stage	Component	Setting applied	Resulting tag	Description
Start	Account Default	English	English	As the call enters the system, VCC opens a data record and applies the global Account default setting (Priority 3).
Entry	Inbound Line (Paris)	French	French (Priority 2 overrides 3)	The system identifies the specific Paris phone number dialed. This Priority 2 setting immediately replaces the Account default.
Routing	IVR Applet	French-Canadian	French-Canadian (Priority 1 overrides 2)	The customer selects "Canada" in the IVR. The Set Data Source applet triggers a Priority 1 override, which is the highest authority.
End	Final Value	N/A	French-Canadian	At the moment of hang-up, the system "locks" the most recent Priority 1 value into the final analytics report.

Inbound line and outbound interactions

The source of an interaction determines how the initial Language and Profile tags are applied before the interaction plan takes over.

Interaction type	Trigger point	Description	Override behavior
Inbound	Inbound Line	The phone number or digital entry point used by the customer.	If the line has an associated language/profile, it overrides the Account Default.
Outbound	Callback Number	The Caller ID selected by the agent when initiating the call.	If the callback number has an associated language/profile, it overrides the Account Default.

Integrated insights

Conversation Analyzer data is seamlessly integrated into your broader reporting ecosystem.

In Historical Analytics, you can perform high-level trend analysis and monitor agent performance across 100% of your interactions. For more information on reporting and trend analysis, see Conversation Analyzer in Historical Analytics.
If you use Salesforce, transcription data, sentiment, and categories are synced directly to your CRM records, providing a unified view of the customer journey and enabling advanced reporting within Salesforce dashboards. For more information on CRM integration, see Conversation Analyzer in Salesforce.