Introduction

The “Sentiment analysis” component helps you identify the emotional tone behind a text (and, in voice scenarios, behind the audio transcription) using Artificial Intelligence. It is an automatic process that requires no training or human intervention, and it allows you to enrich your flows with decisions based on the customer's emotional state (for example: escalating to an agent if the sentiment is negative).

This component is found in the “Cognitive” group and is available in Voice, Interaction, Chat, and CRM + Webhook routing.

Configuration

1. Open the configuration panel by double-left-clicking on the component.
2. In the component's configuration panel, locate its fields:
   • “Variable with the text to analyze”: Here you must type the variable that contains the text on which the sentiment will be calculated.
     • Chat Example: a variable where you already have the customer's message.
     • Voice Example: an output variable from an ASR component (first you convert voice to text and then analyze that text).
     • Important: this component analyzes text. In voice, the “audio” must be previously converted to text using ASR.
   • “Variable with the sentiment”: You will see a predefined variable ($sentiment). This variable cannot be changed, and the result of the analysis will be saved there.
   • “Sentiment analysis value range”: The range table used by wolkvox to classify sentiment is displayed. Based on the visible configuration, the result is interpreted as follows:
     • [-10, -6.25) → Very negative
     • [-6.25, -2.5) → Negative
     • [-2.5, 2.5) → Neutral
     • [2.5, 6.25) → Positive
     • [6.25, 10) → Very positive
   • This value/classification becomes available in $sentiment for use in conditions, decisions, logs, reporting, or subsequent messages within the flow.
   • Click on “Save” to apply the component changes.