All Collections
Understanding sentiment analysis
How does sentiment analysis work?
How does sentiment analysis work?

Welcome to the magic of Machine Learning!

Julian Cook avatar
Written by Julian Cook
Updated over a week ago

Sentiment analysis uses Natural Language Processing (NLP) algorithms to break down a block of text into its component parts. It looks at the structure of words and sentences to see how much emotion is being expressed and whether that emotion is positive, negative or neither.

NLP algorithms rely on Machine Learning - meaning that computers have been ‘trained’ on incredibly large datasets to ‘learn’ what is positive or negative in the context of human communication. While there are some subtleties to human communication that computers struggle with, computers benefit from having a far better ‘memory’ than humans, which makes sentiment analysis highly accurate.

Howamigoing has taken some of Google’s NLP algorithms and trained them to determine, in the context of peer-to-peer and pulse feedback, which comments are:

  • Positive (i.e. the overall feeling within the response is clearly positive)

  • Negative (i.e. the overall feeling within the response is clearly negative)

  • Neutral (i.e. the response is not overwhelmingly positive or negative)

If we take the question “How would you describe your company’s culture?”, then an example of a clearly positive sentiment response would be “Caring for others and helping them succeed in life”.

An example of a clearly negative sentiment response would be “Could be better - lack of communication.”

And an example of a response that is not overwhelmingly positive or negative would be “Okay, but caring more about management and money.”

Did this answer your question?