Configuring sentiment settings in text predictions
Define a sentiment score range to specify the type of sentiment feedback that you receive: positive, negative, or neutral.
You define neutral sentiment within the available score range ( -1 to 1 ). Sentiments with a higher score than the neutral range are positive, and sentiments with a lower score are negative. This setting is helpful when you need to comply with your business requirements and precisely adjust the sentiment ranges. For example, narrowing the negative score range helps to identify the most critical text-based content, such as emails and chat messages.
- Open the text prediction:
- In the navigation pane of App Studio, click Channels.
- In the Current channel interfaces section, click the icon that represents a channel for which you want to configure the text prediction.
- On the channel configuration page, click the Behavior tab, and then click Open text prediction.
- In the Prediction workspace, click the Settings tab.
- In the Sentiment settings section, enter a minimum and
maximum score to define the score range for the neutral sentiment, or keep the
default values -0.25 and
0.25.
Sentiment settings in a text prediction Note: Do not define the neutral sentiment score range as -1 to 0 or 0 to 1 because these ranges interfere with sentiment analysis of input texts. The first score range excludes negative sentiment from sentiment analysis; the second score range excludes positive sentiment. - Click Save.
To understand this configuration, analyze the following text with the default sentiment score values: Your company provides very good service. Still, the prices are too high. I have a neutral opinion about you.
The first sentence has positive sentiment, the second negative, the last one neutral. The overall sentiment for the whole text is neutral because the sentiment score equals 0.03, which is in the neutral sentiment score range ( -0.25 to 0.25 ).
Previous topic Configuring topic settings in text predictions Next topic Configuring preprocessing models in text predictions