Sentiment Analysis: What is it?
On a bright sunny afternoon, I wanted to hangout with a couple of friends and so I was on my laptop searching for restaurants with good ratings and reviews. As usual, google search results gave lots of results. Though you could filter the result based on the ratings but I was more interested in the reviews (written by customers). Consider the review below.
The restaurant was rated 5 star (yup that’s awesome) with impressive review. But consider the review below after some months. Things have changed and from 5 stars it dropped to 2 stars. well we could argue that one person’re review doesn’t matter a lot. Hmnn maybe you want to think about it again. Someone’s bad review might affect those that intend to come and check it out.
What does this say for a business? The fact that you got a good review and rating when starting up is not a guarantee that it would continue like that. I could give a possible reason for this downward review. It could be due to change in staffing ( chef or attendant). So technically, I ignored the 5 star ratings and went straight to the comments of the customers. I decided I won’t go there. Thankfully, there is a way to make analysis of the opinion of individuals (customers) interaction of your services. It would help in adjusting and figure out how to improve for good customer satisfaction. I mean every business is profit oriented “innit”?
Sentiment analysis or opinion mining is a way of identifying and extracting subjective information to determine the attitude of a speaker or writer with respect to a particular topic. It uses Natural Language Processing and machine learning algorithms to be able to detect the polarity of a given text. It’s more of discovering positive or negative feedback from customers or reviewers of product. Its like a measurement to see how the opinion of users has changed over a business product or services.
There are various ways by which sentiment analysis or classification can be done. The amount of data you have is an important determining factor for choosing which machine learning algorithm to train a sentiment classification model. Recently, there has been advances on using Deep Learning for classification task but this works well with enough (sorry large ) amount of data.
Let’s see how to implement a sentiment classification model in the next article here
A project I am working on which would be uploaded later on is a real time sentiment analysis of various Nigeria presidential aspirant for the coming 2019 general election. Can’t wait.
Originally published at www.jerryfadugba.com on February 27, 2018.