Sentiment Analysis

Sentiment Analysis detects the underlying emotion in customer calls and classifies it as positive, negative or neutral.

View sentiment score

To view a sentiment score, open the call details and click Transcript tab.

In the Sentiment score panel, you can see 3 numerical values:

• Overall sentiment score for the call
• Agent-side sentiment score
• Customer-side sentiment score

A numerical value is in range from -100 (negative) to +100 (positive).

The sentiment is color coded (red for negative and green for positive) and is represented with emoji.

Besides a numerical value, the explanation of the score is provided as an evidence. This explanation helps the reviewer to understand why the conversation was classified that way.

Review phrases with negative sentiment

When a call is classified as negative, the phrases that most represent the negative emotions are highlighted with red color in the transcript.

Such a highlighting allows you to quickly spot problem moments in the conversation.

Show Sentiment Score column

The sentiment score value can be displayed as a column in the Recordings page.

To display the Sentiment Score column in the Recordings page, navigate to Administration > Customization > Field Visibility in the admin portal.

Click Edit next to the corresponding layout (we recommend editing at least Recordings - All Calls and Recordings - Advanced Search layouts).

In the Edit Layout page, add the Sentiment score column to the Visible columns list.

With such settings, the Sentiment Score column will be display in the Recordings page.

Search calls by sentiment score

You can use the Sentiment Score in search criteria. For example, to locate all calls that have a sentiment score lower than 0, navigate to Recordings > Advanced Search, select the Call - Sentiment Score as a parameter, Lower than as an operator and 0 as a value. Press Search button.

Additionally, the following attributes are available as a search criteria:

• Sentiment Label. The labels are Very Positive (score 60 to 100), Positive (score 20 to 60), Neutral (score -20 to +20), Negative (score -60 to -20) and Very Negative (score -100 to -60).
• Sentiment Details. A full text search in the sentiment explanation field.
• Sentiment Agent Label. A sentiment label for the agent side
• Sentiment Agent Score. A numerical score for the agent side
• Sentiment Customer Label. A sentiment label for the customer side
• Sentiment Customer Score. A numerical score for the customer side

Use sentiment score in reports

You can include the sentiment score column in various Call Summary and Call Detail reports.

The following screenshot demonstrate the Call Summary report with total calls in each sentiment category.

Sentiment dashboard

You can navigate to Dashboard > Sentiment to see a high-level view of the sentiment scores for a period of time.

If the option Compare to previous period is checked, then a difference is shown in each metric between the current period and the previous one.

The dashboard support top-down analysis. At the top level, you can see sentiment scores for all calls in your organization, then you can go down to the group level, and finally to the agent level.

You can use the search panel in at the top of the dashboard to filter calls as necessary, for example, you can filter data to inbound, duration more than 5 minutes and with the assigned topic "Loss of Business".

Sentiment heatmap

Navigate to Dashboard > Sentiment Heatmap to see a heatmap of sentiment score on a timeline per group or user level.

Such a heatmap dashboard allows you to identify spikes in sentiment scores or call volumes.

A color in the heatmap diagram depends on an overall volume of calls as well as a total calls in negative or positive category (depending on selection).

Note

The sentiment heatmap diagram shows only last 30 days of data for the selected period.

If you would like to see data for the past period, change the date range in the filters criteria at the top of the dashboard.

To see the call metrics for a specific day in the Sentiment Heatmap, hover with your mouse over the certain user/day cell.

Customize sentiment score prompt

Sentiment Analysis engine uses Large Language Mode (or Generative AI) to analyze every call and classify into positive, negative or neutral category.

With the MiaRec platform, you are able to customize the prompt that is sent to AI during sentiment scoring.

Particularly, you can change the definitions of "negative calls" and "positive calls".

The default definitions are:

Score the conversation as positive when the customer was satisfied with the service, all their questions have been answered, and all issues resolved during the call.

Score the conversation as negative when customer remained upset and frustrated at the end of the call because the issues have not been resolved due to agent or company fault.

By changing these definitions, you can improve sentiment scoring in your specific scenarios.

To edit the Prompt for Sentiment score, navigate to Administration > Speech Analytics > Generative AI > AI Insights. Click Edit next to the Sentiment score insight.

Note

You need to have the appropriate admin credentials to edit a prompt for the Call Summary in your organization.

You can test your new prompt in AI Playground on your own calls to make sure it works as expected.

To test the prompt, click Test button on the Sentiment Score AI Insight page.