Sentiment analysis

The sentiment analysis feature analyzes messages during a conversation between a human agent and an end user to determine emotional intent. Conversational Insights only measures the sentiments of end users in contact center conversations. Sentiment is represented by a numerical score and a corresponding text label. The score lies in the range (-1, 1), neither values included.

Turn-level sentiment

Conversational Insights calculates the sentiment scores for individual user turns within a conversation as follows:

Score	Magnitude	Fine-grained sentiment	Sentiment
-1	1.0	negative	Negative
-0.5	0.5	somewhat_negative	Negative
-0.25	0.75	mixed_more_negative	Negative
0	0	neutral	Neutral
0	0.75	mixed_balanced	Neutral
0.25	0.75	mixed_more_positive	Positive
0.5	0.5	somewhat_positive	Positive
1	1.0	positive	Positive

View turn-level sentiments scores

View the sentiment scores for each turn within a specific conversation as follows:

Conversation-level sentiment

Conversational Insights calculates the sentiment scores over entire conversations as follows:

• Positive: Greater than or equal to 0.5
• Neutral: Between -0.5 and 0.5
• Negative: Less than or equal to -0.5

View conversation-level sentiment scores

For each agent in your contact center, you can view the sentiment score for each of their conversations or the average sentiment score over all their conversations.

You can also view sentiment scores for individual conversations from your full contact center conversation list.

What's next

• For more details on interpreting sentiment analysis results, see the Natural Language sentiment analysis documentation.
• For examples, see the Agent Assist sentiment analysis documentation.