Intro

We’ve been asked many times about how AI is applied to do this, how AI is used in that sector, how to bring AI to existing products. For this reason, we’ve decided to write a series of posts. To write something useful, we’ll keep them short, clear and provide few relevant references.

Don’t hesitate to suggest topics of interest :)

What is Sentiment Analysis?

Sentiment Analysis aims to understand how someone feels about something.

selective focus photography of woman wearing black cold-shoulder shirt using megaphone during daytime
Photo by Clem Onojeghuo / Unsplash

Where is it used?

Sentiment Analysis is currently used in many areas here we have few:

  • Brand sentiment: aims to understand how people feels about a given brand.
  • Customer satisfaction: measures the happiness of a customer about a product or service received.
  • Customer service: monitors a conversation or call to assess the overall satisfaction and alerts the right person when the customer is about to scale the issue.
  • Product recommendation: if you like a novel or song, it’s likely you like similar novels or songs.
woman wearing white headphones
Photo by bruce mars / Unsplash

How does it work?

The data from customers can be collected in so many ways, from likes and dislikes to a post, to reviews on Amazon or podcasts on YouTube.

In any of these forms of communication, there is a topic and data to be analyzed surrounded by clicks, product views, opened links, comments, reviews, messages, emails. Then it’s just a matter of mining them to evaluate the sentiment.

Hands-on

Let’s see a simple example in which we use a dictionary of words scored by sentiment. In it we can find many scores here we have some: wow (4 points), interesting (2 points), weak (-2), worse (-3 points). In this fashion, positive words increase the score and negative words reduce it.

Let’s pick 2 real sentences from iPhone X reviews:

  • “Great sound from the stereo speaker. Camera produces AMAZING pictures”: Score: +3 (great) +4 (amazing) = 7
  • “Android ripoff, but in a bad way, with very confusing gestures.” Score: -3 (bad) -2 (confusing) = -5

This first review scores 7 points whereas the latter scores -5. This means that the first one is more positive than the other. Also tells us that the first review is positive with a score above 0 and the other negative as it’s score is below 0.

That’s how in a simple way a computer can tell apart a positive review from a negative one.

That’s all

Hope you like it! Comments will be always welcome :)