All Collections
Frequently Asked Questions
What is Sentiment and how is Sentiment calculated?
What is Sentiment and how is Sentiment calculated?

Everything you need to know about what Sentiment is and how Sentiment is calculated in Kapiche.

Ryan Stuart avatar
Written by Ryan Stuart
Updated over a week ago

What is Sentiment Analysis?

Sentiment Analysis is the process of using Natural Language Processing (NLP) to automatically detect the sentiment in a piece of text that someone has written

(e.g. this could be in the form of a survey response, an email, a tweet, etc.!)

Because humans often love to express emotions in their writing, Sentiment Analysis has become a powerful way for businesses to monitor how their customers feel about their brand / experience and how those feelings change over time.

How does Kapiche calculate Sentiment?

Kapiche uses Plumeria - our own sentiment algorithm - to determine the sentiment someone has expressed in their writing at both an overall and at a sentence level!

Plumeria has been trained to calculate the probability of sentiment (positive, negative, mixed and neutral) in someone's writing based on the types of words they have used and how those words interact with one another throughout the text.

For example, Plumeria knows from past experience (that's the training component!) that when "love" appears alongside "absolutely" (e.g. "I absolutely love Kapiche!") then the sentiment is almost guaranteed to be Positive.

On the flipside, Plumeria also knows that if the word "don't" appears between "absolutely" and "love" (e.g. "I absolutely don't love Kapiche"), that the sentiment would instead be Negative!

And because Plumeria processes text at an overall and at a sentence level, it's able to identify when someone is expressing more than one emotion (e.g. "The staff were amazing! Unfortunately my food was cold, though," would be detected as having Mixed sentiment).

How does Kapiche train Plumeria?

The first element of teaching Plumeria how to accurately detect Sentiment was to feed it extremely large volumes of data sets similar to yours that have had the Sentiment classified by a human - this is what's known as "Training Data".

From there, we use machine learning algorithms that enable Plumeria to continuously learn and evolve over time (e.g. the more data that Plumeria analyzes, the smarter and more accurate it becomes!)

What's really exciting about this approach is that eventually you will be able to help train Plumeria yourselves using your own data - or more specifically - using the unique language your customers use to talk to you!

Questions? πŸ€”

If you have any questions about Sentiment analysis (or anything else!) you can get in touch with us any time by hitting the blue chat button to your right πŸ‘‰

Did this answer your question?