Get a Sense of How Your Customers and Prospects Feel

Text information such as product reviews and social media posts can be a mine of information for your business. Use the Detect Sentiment transformation in a Data Prep (Beta) recipe to quickly bucket that information into sentiment categories: positive, negative, and neutral. For example, detect the sentiment of survey responses to evaluate how customers feel about your product support. If more than a certain percentage—say 30%—of the comments are negative, escalate the feedback to support management.

Where: This change applies to Einstein Analytics in Lightning Experience and Salesforce Classic. Einstein Analytics is available in Developer Edition and for an extra cost in Enterprise, Performance, and Unlimited editions.

How: If you haven’t already done so, enable Data Prep (Beta) in your org. For more information, see Prepare Data with the Next Generation of Data Prep (Beta).

In a Transform node of a Data Prep (Beta) recipe, select the dimension column in Preview, and then click the Detect Sentiment transformation.The Detect Sentiment button appears in the Transformations toolbar when you select a dimension column.

When you run the recipe, Analytics writes the results to a new column. If needed, edit the column label and click Apply.The Detect Sentiment screen allows you to change the label of the new column that will store the sentiments.
Note

Note

The Show Results In field is grayed out because this transformation can only write the sentiments to a new column and keep the original dimension column. Unlike other transformations, you can’t select the other options to write the results to a new column and discard the original column, or overwrite the original column.

Preview shows “Sentiment TBD” in the new column. The sentiment column shows "Sentiment TBD" until you run the recipe.

This transformation supports English text only. It processes non-English text as English and ignores images, including emojis. The transformation returns a null if the sentiment can’t be determined, such as when the input value is null. Typically, the transformation generates the correct sentiments, but the results can vary based on the data. For example, text without sentiment (such as IDs, nouns, addresses, and alphanumeric values) can return unexpected results.