Let me introduce you to my friend Anne. Back in the day She started a small clothing business. She used to sell these to her relatives and friends. She got suggestions and feedback directly from them.

She kept on with the trends and she grew her business. She got customer feedback verbally, through Emails and Surveys.

Later into the future, her business became an international brand. She expanded her business more and her clothing was distributed to major department stores across the globe.

So far she made amendments to her business by analyzing her customers sentiments directly. Since she’s having a large number of customers across the world, obtaining direct feedback and analyzing them manually would become a hectic process.

How does she use the newly available methods to obtain and analyse her customer opinions? This is where Sentiment Analysis with Artificial Intelligence comes into play.

In this post, I would like to share What sentiment analysis is, how we analyse people’s sentiments using artificial intelligence, what are the benefits of sentiment analysis to a business. In general, the term sentiment analysis nowadays is defined as,

Sentiment Analysis is a process which identifies a piece of text and catalog them based on the tone conveyed by it.

Basically the goal of sentiment analysis is to know a user or audience opinion on target object by analyzing a large amount of text from different sources. When you want to buy something on the internet, if you are a bit of a person who knows how the internet works, you open up the product pages and look for other buyers comments and reviews of these products.

Nowadays, peoples opinions on certain products have gone so far. You will look for social media like Facebook and Twitter, reputed forums such as Reddit and also popular product review pages such as cnet in order to evaluate if your product is either valuable or not.

How do we collect these raw information?

These Social Media platforms , popular blogs and forums exposes their developer components to software developers to get a hold of these customer opinions.

If you want to gather more information throughout the web, there are web crawlers and web scrapers which automatically browses the internet and gather latest information based on your provided tags and keywords.

Now that we have a large amount of unorganized information from various sources, we need to polish them up to be ready for processing. Here we use data cleansing technologies which are powered by deep learning / AI to remove all the unwanted text from web resources such as symbols, timestamps, advertisements, dummy texts etc.

Processing the collected raw data

We now use this clean information to process and gather the opinions. In AI, we call this Natural Language Processing, which is an artificial intelligence that reads out the information provided and put together some meaningful outcomes.

In sentiment analysis, we categorize these meaningful outcomes to Subjectivity and Polarity.

In simple terms, subjectivity is what the item or the object or the person that is being discussed. Lets see a small example

“I bought the Apple iPhone XR a month ago. The screen quality is so bad. My other Samsung phone is much better than this iPhone. It is also so expensive”

Here,IT means the iPhone XR

“I bought the Apple iPhone XR a month ago. The screen quality is so bad. My other Samsung phone is much better than this iPhone. It is also highly customizable”

Here, IT means the Samsung.

Polarity term refers classifying the sentences based on the opinion. Lets go again with examples.

“That movie is amazing. The ending was the best so far”

So it implies that the user enjoyed the movie. It has a positive polarity

“The movie has a poor storyline. It’s too childish and boring”

Here the user has a negative polarity in his opinion

“I watched this movie because it’s meant to be one of the best so far”

Here the sentence can imply either negative or positive. In such cases, we name it as neutral polarity.

There are more cases in sentences of how people express their opinions. They compare products with one another, sometimes they use metaphors, some comments are sarcastic so on and so forth. Therefore, in order to identify the actual context and the sentiment of the users, we have to use different NLP algorithms.

These algorithms can be, Rule Based or Automatic. Or sometimes they may have both.

In rule based algorithms, we manually set up different rules so that the AI can classify information based on them. For example we define a set of positive words and negative words and look for such word counts and determine the sentiment.

Automatic approaches highly relies on machine learning algorithms. In short we use training data with labels to train a machine learning model. The trained model is then used to identify the trained patterns and predict the labels of any other inputs.

What are the benefits of sentiment analysis in businesses?

Let’s get back to Anne’s international clothing brand. So far she has collected feedback and reviews to improve her products. Using sentiment analysis she can have deeper insights to improve her products. For example, she can now see that people love the colors and the designs of new dresses. But the old designs were comfortable to wear. Now she can use these opinions on her newer designs to be better as well as comfortable.

She also has access to how people’s sentiments are given with a timeline. For example, she can determine the public opinion on this years Christmas items compared with the last years sentiment. Since these data are calculated automatically, she now has a room to fix timely mistakes as soon as possible.

Her clothing products now travel globally. Sentiment analysis tools can identify from where people use her products mostly. Where the customers are with strong opinions. With this information, she can increase the the number of products distributed to such customers and get more feedback on her products.

Even Though she manages her marketing and use communication channels effectively, sentiment analysis tools determine that in some channels people share their opinions more than the rest. For example, the tweets shared about her products are much higher than the comments shared in Facebook. This way she can improve how she communicate with different customers in different channels.

One of the most important challenges with larger companies is prioritizing customer service issues. Using sentiment analysis tools, she can now determine what are the critical ones that needs priority and top them up in the queue. This way she can quickly access negative feedback and convert them to positive.

A Real Case-Study

Let’s look at real world example of how sentiments of people makes an impact on a certain product.

If anybody knows this game No Man’s Sky, it’s one of the unique games made so far. It had a huge hype among gamers with the release of their teaser trailer. It had vast amount of space exploration, very beautiful scenes and animations and the hype was real. Time went on without any release updates of them and the hype was decreasing slowly.

One day they released a statement that the game was going to be delayed. Optimistic people reviewed that the game is being polished to be better so the product gained a bit of positive sentiment.

Then they released the game, and the sentiment went downhill. The game was nowhere near the expectations. No pretty animations, no multiplayer, its laggy, its buggy and peoples expectations were not met.

However, the company identified what went wrong and they released minor updates in order to fix them. But they didn’t help that much.

The company went full silent, kept a couple of months and released a new update. They called it the “foundation update”. This contained most of the features users wanted. Even the features that they initially not thought of. As soon as the release of the game, the sentiment of it started going towards positive.

The following image depicts the real world information wrapped to a sentiment graph along with the timeline of events I mentioned above.(Source) The data was captured from Reddit data by Guillaume Couture-Piché, a data visualization passionate. Check his blog for more amazing posts about data science and visualization.

As you can see this is how the hype went downhill with their release, and how they managed to get back on feet with their new update.

I believe that sentiment analysis is an essential tool for improving your business reputation online. It takes away the traditional methods by analyzing large quantities of data and gives you actionable figures and facts that you can address strategically and tactically.

If you are an owner of a Business, or hoping to be one day, I believe that you too as the decision makers would make use of these new tools to take your businesses throughout the world!