Real-Time Speech Analytics

Real-Time Speech Analytics (RTSA) detects specific words or phrases spoken during captured interactions. The system can send alerts or notifications to employees when detection occurs.

Use this feature to provide guidance to employees when particular topics of interest are discussed on recorded interactions. For example, if a customer speaks the phrase "cancel my account" to an employee on an interaction, the Real-Time Speech Analytics feature could display a webpage to the employee with information about handling account cancellations.

This feature is useful in other scenarios where employees need immediate awareness of topics discussed in interactions. For example, the feature could display a desktop popup to a particular manager when the phrase "service is out" is spoken on an interaction. In another scenario, the feature could send an email message to a manager when an employee does not answer a interaction with the appropriate greeting.

The Real-Time Speech Analytics feature can also take other actions when specific words or phrases are detected on an interaction. Some of these actions include:

• Tag interaction metadata with a user-defined text string when specific words or phrases are spoken on an interaction. Users can enter this text string in a search feature to find and analyze interactions in which specific words or phrases were spoken.

• Assign to a case any interactions in which specific words or phrases are spoken. A user can access this case to play back the interactions.

Real-Time Speech Analytics configuration

Real-Time Speech Analytics can be deployed with or without Speech Analytics. In either case, you follow a specific workflow to configure the Real-Time Speech Analytics feature to detect when specific words or phrases are spoken on interactions. This workflow involves configuring the following areas:

• Recording rules - You create recording rules to select the interactions on which the Real-Time Speech Analytics feature detects words and phrases.

• Speech Analytics Real-time Categories

  • If you have Speech Analytics, in addition to the default set of predefined Real-time Categories, you can also create user-defined Real-time Categories with the terms and phrases the system detects.

  • If you do not have Speech Analytics, you can use the default set of predefined Real-time Categories to detect the terms and phrases.

  You associate these Real-time Categories with a Recorder Analytics Rule.

• Real-Time Speech Analytics Models - For deployments with or without Speech Analytics, if you are using the local Real-Time Speech Analytics Engine to transcribe interactions, upload a Real-Time Speech Analytics Model (a language model) to the system and activate the model. The model determines the language and vocabulary in which the Real-Time Speech Analytics Engine detects words or phrases.

• Organization or system alerts - You configure these alerts to notify employees that the words or phrases were detected on an interaction. The notification can take any of these forms:

  • Email - An email message is sent to particular employees when words or phrases are spoken on interactions.

  • Desktop popup - A desktop popup is displayed on an employee computer when words or phrases are spoken on interactions.

  • Work Assist notification - A message is displayed in the Work Assist application on the agent desktop and on the Work Assist web page.

  • Desktop Process and Analytics (DPA) action - A DPA action executes when words or phrases are spoken on interactions. For example, a third-party application could start on an employee desktop or a webpage could display when words or phrases are detected on interactions.

• DPA Triggers - You can configure DPA triggers to combine events and data from the employee screens with recorder analytics events. For example, you can configure DPA to open a browser with a specific URL to a knowledge base article when an employee accesses a particular screen in an application.

• Custom attributes - You map a custom data attribute to a Custom Data field. This configuration enables the feature to tag the interaction metadata with a user-defined text string.

• Recorder Analytics Rule - You create a Recorder Analytics Rule to enable the Real-Time Speech Analytics feature and define the parameters with which the feature operates. These parameters include the recording rules used to select interactions for analysis, Real-time Categories used to detect words and phrases, and the engine used to perform real-time time detection of words and phrases:

  • Real-Time Speech Analytics Engine: The Real-Time Speech Analytics Engine performs analytics processing locally on the same Recorder on which interactions are recorded. The Real-Time Speech Analytics Engineis designed for high performance during live processing of audio streams, with a marginal tradeoff in accuracy compared to an offline transcription engine. Transcription accuracy can be improved by refining category definitions to enhance the precision of category hits, and by tuning the language model.

  • Real-time Linguistics (remote) Engine: The Real-time Linguistics (remote) Engine performs the same functions as the Real-Time Speech Analytics Engine, except with this engine, the analytics processing is performed in the cloud rather than on the Recorder on which the interaction is recorded. This approach conserves processing resources on the Recorder on which the interaction is recorded. The Real-time Linguistics (remote) Engine provides enhanced transcription accuracy and precision of category hits compared to the local Real-Time Speech Analytics Engine.

Real-Time Speech Analytics analysis

If you tag the interaction metadata with Custom Data (such as a user-defined text string), you can use the Interactions application to search for interactions tagged with the text string. When you replay the interaction in the playback tool, the words or phrases detected during the interaction appear as annotations in the tool. You can click these annotations to go directly to the point in the interaction at which the words or phrases were spoken. You can also use the Risk Management search feature to search for tagged interactions.

You can run reports to analyze the notifications data. The Recorded Contact Volume report and the Recorded Contacts per Agent/Group reports are useful for analyzing notification data and monitoring notifications rates.

Sensitive Data Masking

If Sensitive Data Masking is configured in your system, the system prevents the playback of sensitive information in an interaction. You can configure Sensitive Data Masking by using an RTSA rule.

Based on the RTSA rule, the system detects that four or more numbers are spoken in a sequence of six words, and appends Data Masking Start and Data Masking End tags to the interaction to mark the start and end of a portion containing sensitive information. During playback, the player then masks the portions of the interaction that are marked with the tags.

Related information 

Real-Time Speech Analytics configuration (Real-Time Speech Analytics Setup Guide)

Player user guide