Sentiments FAQ

What is sentiment analysis?

Sentiment analysis identifies the polarity of text (Negative, Positive or neutral). Read more on What is sentiment analysis.

How to run sentiment analysis on my data?

The easiest way to access sentiment analysis is through the AI-Clustering workflow. More details are available on Run Sentiment.

What sentiment information does Relevance AI provide?

After running the sentiment analysis, you will notice, a new column/field is added to your dataset. Under this field exist two different values, sentiment label and sentiment score.

  • Score: The confidence score on the decision made by AI on the polarity of the text
  • Sentiment: A string field containing positive, negative or neutral

How to visualize sentiment

See our full guide on sentiment overview on Explorer dashboard.

What does sentiment ranking mean on Explorer?

On the Explorer category view, categories can be sorted and ranked based on different parameters. You can read more on the Rank by Sentiment page.

When it doesn't make sense / where caution should be applied in interpreting extracted sentiment

Sentiment analysis results add more value when people are asked about something in general. For example, if the question is "What was your best experience about ...?", responses are likely to be all positive. Whereas when the question is "What feedback can you provide about ...?", responses are expected to be both positive and negative. Hence, more value in sentiment analysis.

Whether scores can be used quantitatively

Yes! Sentiment scores are numeric values between -1 an 1, with -1 representing pure negative and 1 pure positive. Under setting, you can specify the range for neutral.

How to export scores and What should I expect

To export sentiment analysis results to a CSV file, please follow export on Explorer. Just make sure that you have followed through How to run sentiment analysis and added a sentiment metric. On export, you can

  • download overall sentiment scores over categories
  • download sentiment labels and scores per entry per category