AI - Detect Sentiment

Many times it can be useful to automatically detect if your users are feeling happy or upset based on their messages. As a common example, if you ask your users to provide detail about a customer service issue before beginning a live chat, it can be helpful to determine if the user is angry, and possibly prioritize users who are having trouble or marking those chats for later review.

You can easily detect sentiment in Flow XO using our Detect Sentiment task. This task is part of our AI package of integrations.

To configure the detect sentiment action, all you need to do is provide the text you want to detect the sentiment of. Usually this will come from a user message.

TText - the user input or other text you want to know the sentiment of

Language Code - optionally, provide the language code of the language the message is written in, if you know it. If not, the language will automatically be detected. This setting is only necessary if you are not getting good results for your language and may sometimes improve results in those cases, but is usually not necessary.

And that's it. The outputs of this task are simple:

Sentiment - Either "Positive", "Negative" or "Neutral"

Confidence - a score that shows how confident the model is about its guess. It's a number that ranges from 0-1, with 0 being the model has completely guessed and 1 being the model is 100% certain of its answer.

Here's a few examples of this step in action:

And that's it! You can use the resulting sentiment to categorize messages, route a flow differently for negative or positive sentiments, or to tailor your response based on the sentiment of the user.

Happy flowing!