How do we calculate sentiment?

Updated 2 weeks ago by Sofia Quintero

Sentiment Analysis

Address customer issues quickly, and find great quotes from your users. With your documents analysed for their sentiment, you can filter your feedback in the most simple way; the good stuff, and the bad.

You will notice these icons below appearing next to your feedback in NomNom.

How do we calculate sentiment?

Documents from your integrations, historic and new, get analysed for sentiment as they are processed by NomNom.

We run a sentiment analysis algorithm to distinguish if your document is Positive, Negative or Neutral.

This is a model that we have built using a large data set of human classified data which have Positive, Negative and Neutral language.

We use learnings from the model we have built to give each bit of feedback from your integrated sources Positive or Negative sentiment. If the bit of feedback doesn't follow a Positive or Negative pattern, then it is given Neutral sentiment.

NomNom applies sentiment automatically to integrated sources, but not to documents such as Notes, Spreadsheets or Google docs. You are able to apply sentiment to these sources manually.


NomNom converts any systems which assign an NPS score into sentiment. This can come from NPS platforms, and any NPS scores you are importing from surveys.

NPS Scale Conversion
  • Postive => 9-10
  • Neutral => 7-8
  • Negative => 0-6


You may find that you have a document in your account with positive sentiment, which has been labeled as Negative, or vice versa.

This is normally down to some complexities in the types of language your customers are using.

The accuracy levels increase over time as the algorithm learns about your data, especially when you manually change the sentiment of your documents. The more you interact with NomNom the more it learns.

If you want to learn more about sentiment analysis technology you can check this useful video:

Text By the Bay 2015: Richard Socher, Deep Learning for Natural Language Processing

You can improve accuracy by manually changing sentiment. By manually changing the sentiment of a piece of feedback, you are training the machine to understand the type language your customers are using when interacting with you.

We are constantly working on improving our algorithm to increase the accuracy when applying sentiment to your documents!

How Did We Do?