Emotion Analysis from Text

Gain insights, act quickly and reduce customer churn.

Benefits

Identify patterns in customer emotion - Analyse messages to extract information on customer emotions, utilising valuable data in order to gain insights for decision making.

Get Trend insights earlier - Have the ability to act more proactively to customer trends, or in order to take advantage or positive experiences for promotion and insights.

Improve Customer Experience - Understand your customers’ needs better - enabling you to create a more user-centered product that keeps customers on board!

Pain point

We all know that the number of stars in a review doesn’t tell the whole story. Reviews are often very subjective, and a 4-star review can be more scathing and critical than a 2-star review. How could you possibly know how your customers feel if you have no basis for the review?

Solution

The Emotion Analysis from Text workflow extracts details from the reviews in a way that allows businesses to understand the emotion behind any given review and identify trends and commonalities between a range of reviews.

Usage (Input/Output)

Text is consumed as input using a JSON file (find the details in the technical API docs). The input is broken into sentences, which are then analysed for their underlying emotions. The output is then returned in a JSON response in a sentence:emotion pair.

Expectations/Limitations

This model performs best on emotionally heavy text chunks such as reviews of products and services.

The JSON format of the file follows the output of AmazonTranscribe. We will simplify this in a future iteration.