We analyze sentiment and emotions in the content

It is an automatic or semi-automatic study of feelings, impressions or opinions expressed on a given subject. It can consist of a request for the participants to indicate their emotions. However, statements are analyzed more often, especially those published online – then we talk about analyzing emotions in a text or even monitoring them. Of course, when thinking about emotions we count something that expresses them – words, whole texts, emoticons, likes and dislikes, the number of stars next to the opinion.

Monitoring emotions allows to measure the moods of investors and clients. It is also an early warning mechanism against crises. It can show the condition of communication with the client or internal communication in the company (growing anger in e-mails does not serve for good cooperation). It also shows how feelings about a chosen topic change over time, what temperatures are reached by public discussion in social media and more.

Any theory or analysis of emotions considers some set of basic emotions. For example, Ekman (joy, sadness, fear, anger, surprise, revulsion) or Plutchik (joy, trust, fear, surprise, sadness, revulsion, anger, expectation). There are many models and definitions of basic emotions. What they have in common is the belief that such feelings are painted on the faces of people from different cultures in the same way, basic emotions are an immediate reaction to evolutionary important stimuli (so we can share them with animals as well), and all the higher feelings can be described as assemblies of these basic ones.

With our proprietary tools we analyze the content of emotions in texts written in Polish: Facebook posts, newspaper articles, emails, comments from clients. Thanks to this, we measure the temperature and effectiveness of communication, warn against crises, get to know market moods, we can also show what brands and names are best associated with.

Of course, we use two models of analysis of the overtones of the text: bivalent sentiment and 8 basic emotions. The latter is much more interesting and much more accurate in describing the emotional overtones of the content. We also measure sentiment more effectively and accurately – we use a multi-stage intensity scale for negative and positive overtones, as well as the strength of stimulation.

It’s best to collect entries – of the brand itself, comments under its posts, mentions on the profiles of other institutions and private individuals, and then harness artificial intelligence. Of course, you can manually mark posts as positive, negative, containing fear, anger or joy, but it’s better to rely on automatic taggers and classifiers – algorithms created by us.

On our blog, we also write about how to conduct an analysis well, respecting statistics and scientific objectivity.

We write extensively on our blog about 8 basic emotions from the Plutchik model. Each of the variables has a different meaning, and the analyses based on this model allow for better understanding of messages and predicting the reactions or behavior of the authors and recipients of messages.

There aren’t :). As the only one in Poland or maybe in the world, we analyze not only sentiment, but first individual emotions in the content. This is much more than it was available before. You’ll admit that it’s a different quality when you find out – as before – that a given text has a negative sentiment, and when – thanks to Sentimenti – you’ll see precisely whether it’s anger, repulsion or maybe sadness.

Moreover, Sentimenti will show you how the same content is perceived differently by people depending on their age, education, material status, relationship status, having or not children, owners or not animals. Instead of sentiment, true emotion.