Using analytics search with Intelligent Voice

Owned by Gerry Durning
Last updated: Dec 04, 2023 by Zszugyi
5 min read

Overview 

Organizations use Verint Financial Compliance's Analytics Search capabilities to reduce the time effort required to find potential compliance issues within Conversations.  This guide will describe how to:

  • Search within Transcripts
  • Search by Language
  • Search by Sentiment
  • Search by Topics
  • Understand the available columns within the Conversation Search results grid

Search within transcripts

This functionality empowers the user to search within transcribed recordings. 

The search fields support '*' for a starts-with or ends-with match, or " " for an exact match.

Please note, if IV is configured to transcribe multiple languages, then the user can type phrases in the supported language set.

Search Field Name
Description
All of these phrasesWill search across transcribed conversations for transcripts that contain all of the listed phrases in one transcript.
Any of these phrases

Will search across transcribed conversations for transcripts that contain any of the listed phrases in one transcript.

None of these phrases

Will search across transcribed conversations for transcripts that contain none of the listed phrases in one transcript.

You cannot search within transcription summaries.


Language search

This functionality empowers the user to search for use of a specific language within a recorded and transcribed conversation.  The functionality also empowers the user to search for conversations where the recorded user, switched from one language to another.  

To use Language Search, the environment must be configured to transcribe the target language.  Please refer to the ASR model management for more details.
Search Field Name
Description
Language switched during conversationWill search across transcribed conversations where a user has switched to a second language during the conversation.
Language Spoken

Will search across transcribed conversations for transcripts that contain use of a recognized language. Please note, multiple instances of the language spoken search field can be added.

Search within sentiment

This functionality empowers the user to search by sentiment within transcribed recordings. 

Prior to explaining the search capabilities, it is essential to understand the granular sentiment analysis capabilities of the system.  Sentiment analysis is conducted at a fine-grained level, each diarized, speaker separated transcribed text is analyzed and given a sentiment score, within the system this is referred to as a TCU instance.  As sentiment is analyzed and stored at such a granular level, richer sentiment analytics and search capabilities are enabled.

Search Field Name
Description
Sentiment Gradient

An additional filter is available to refine the sentiment shift. This enables the user to define a search filter (more than or less than) shift by Numeric Value, where the numeric value (-99 to +99) represents the sentiment change

Sentiment Gradient

Enables the user to search for conversations where the sentiment has shifted in different directions. For example:

  • Conversations where there has been a significant sentiment shift from negative to positive sentiment.
  • Conversations where there has been a significant sentiment shift from positive to negative sentiment.
  • Conversations where there has been a sentiment shift from negative to positive sentiment.

  • Conversations where there has been a sentiment shift from positive tonegativesentiment.

TCU Instances

Enables the user to search for conversations where there has been a large volume of negative or positive TCU sentiments in a given conversation. The user is able to refine the search to focus on 'Very Negative', or 'Negative', or 'Positive', or 'Verint Positive' instances of sentiment TCU. The user can set the numeric range.

Search within topics

This functionality empowers the user to search by Topic. 

Prior to explaining the search capabilities, it is essential to understand Topics.  Topics are generated for each conversation, topics represent linguistically significant text within a given conversation.  The system automatically identifies key Topics spoken within a conversation, providing a quick summary of the audio file content. 

Search Field Name
Description
Topics

Enables the user to search for conversations that contain specific Topics.

Search results

The Conversation Search includes columns for the enriched speech analytics insights. 

The following columns can be added to the search results.

Search Result Column
Description
Language SwitchedHighlights if a language switch was detected during the conversation.
LanguagesLists the detected languages spoken during the conversation.
Topics

Lists key topics spoked within a conversation, providing a quick summary of the audio file content.

Sentiment Gradient

Provides a graphical display of the sentiment shift during the conversation. The angle and direction of the line reflects the sentiment direction and gradient.

The sentiment gradient maps the changing sentiment (tone or mood) across the entirety of an interaction by calculating the change in the sentiment between the beginning and the end of the conversation for each speaker. Note that a positive sentiment gradient does not necessarily mean that the outcome of the conversation was positive, it only reflects the direction of the change in the sentiment across the call.

The Verba system displays the sentiment gradient on the search result screen as follows:

  • The gradient scores are provided separately for each identified speaker by the Intelligent Voice solution. The Verba system only displays the best and the worst sentiment gradients for the conversation. Note that this means if there are more then 2 speakers in the conversation, the system will not display all sentiment gradients.
  • The sentiment gradient is presented on a scale of 4 categories and displayed as a small line chart:
    • Sentiment has changed very negatively 
    • Sentiment has changed negatively 
    • Sentiment has changed positively 
    • Sentiment has changed very positively 
TCU Sentiment

Lists the granular count of TCU sentiment within the conversation. 

The numbers with the colored background represent the number of TCUs in the respective sentiment category for the recorded conversation.