The comparison: Emotional analysis vs. sentiment analysis

Published on May 23, 2022  - Updated on December 05, 2023

The comparison: Emotional analysis vs. sentiment analysis

When it comes to using an automatic semantic analysis tool, there are two approaches: sentiment analysis and emotional analysis.

Although similar in many ways, there are fundamental differences between them that are important to understand in order to make the right choice.

In this article, we'll compare the two approaches to define which one to choose for your situation. 

What's the difference between sentiment and emotion?

Before we begin our comparison, it's important to discuss the difference between sentiment and emotion, particularly in the context of semantic analysis. 

Let's start with feelings.

The latter are considered opinion markers and can be represented by different linguistic forms (morphological, syntactic, prosodic) and appear as personal or impersonal verbs, adjectives, adverbs, or even certain emojis or hashtags. 

Sentiments can be predominantly negative or positive and are used to express certainty, doubts, agreement, or disagreement on a subject.

Emotion, on the other hand, is an intense, often uncontrollable affective state that prevents us from reacting reasonably and appropriately to the event that provoked it. Emotion is so strong that it tends to overwhelm us and cause us to lose our composure. Unlike sentiments, emotions play a key role in decision-making.

There are 6 so-called "primary" emotions. These are happiness, surprise, fear, sadness, anger, and disgust. These are the emotions that, when used properly in marketing, can influence our perception of the real value of a product or brand, and consequently direct our choice towards one product rather than another.

This is why many brands seek to use emotions in their strategies (also called emotional marketing) to create or reinforce consumer loyalty.

In the rest of this article, we'll compare the semantic analysis of feelings and emotions, so that you can see which one best suits your needs.

1) Sentiment Analysis

If you have ever used an automatic semantic analysis tool, it is very likely that you have already heard of (or even tested) sentiment analysis.

Indeed, most of the specialized tools today offer semantic metrics and notably a sentiment score. The objective is to classify each comment between positive, neutral or negative sentiment with a confidence rate and possibly an intensity rate.

Behind the results, a pre-processing step will first clean and standardize the words. The most common techniques will automatically lowercase, remove punctuation, numbers and empty words, special characters can be converted, the word family analyzed, etc.

Once the pre-processing is complete, the automatic sentiment algorithms can be classified into 3 parts: 

• machine-learning: this approach uses a machine learning technique and various features to build a classifier that can identify the text that expresses a sentiment. Nowadays, machine learning methods are popular because they are easily adaptable to any data. 
• lexicon-based: This method uses a variety of words annotated with a polarity score, to decide the overall evaluation score of a given content. The strongest advantage of this technique is that it does not require any training data, while its weakest point is that a large number of words and expressions are not included in the sentiment lexicons.
• hybrid: The combination of machine learning and lexicon-based approaches to deal with
  sentiment analysis is called hybrid. Although not commonly used, this method generally produces more promising results than the previously mentioned approaches. 

Enormous progress has been made in the natural language processing area, and the best tools now achieve a reliability level of over 75%. 

If the analysis conducted by the company does not require significant depth, then sentiment analysis can provide some benefits.

• Crisis prevention and management on social networks. When a company is in a bad-buzz situation on the networks, it is necessary to act quickly. If an abnormal volume of negative verbatims is detected on a network such as Twitter, the company can take the matter in hand immediately and act more quickly.

• Manage the E-reputation. Often integrated in Social Listening software, sentiment analysis allows to monitor the e-reputation of a company on social networks. By tracking and filtering by sentiment what is said about your company, it allows to have a greater knowledge about the general opinion about the brand.

But there are some limitations and drawbacks in sentiment analysis: 

• sentiment is not emotion... Sentiment analysis does not offer enough depth or capability to be useful for survey feedback or web review analysis. It is not a complete replacement for reading survey responses as there are useful nuances in the comments themselves. 

• sarcasm/irony is not accurately rated. (And if so, should it be a positive or negative sentiment!?). 

• another drawback is that the ambiguity of natural language can confuse NLP algorithms, which is why the neutral part is often the main part (so a lot of data is needed to start detecting irritating points or building a customer journey map).

That's why, more recently, researchers and linguists have challenged existing sentiment analysis for more depth and more actionable scoring. They built the first emotion analysis tools that first spread to the best customer experience management platforms. 

2) Emotional analysis

If you want to have a global understanding of the customer experience and customer journeys, then emotion analysis is what you need.

But how does it work?

The tool will analyze the comment, decide which emotion is the most appropriate for each sentence as well as its emotional intensity.

To achieve this result, millions of sentences, words, etc. have been integrated into an emotional dictionary and coupled with layers of interpretation, which makes it possible to interpret all the signals used in the written language. For example, at Q°emotion, we have developed a multilingual emotional dictionary with more than 50 million words and phrases to perform this analysis.

Capitalization, negative forms, punctuation, emojis, inversion, verb tense and forms will be useful to better detect the emotion of the comment and limit the number of errors. 

By integrating emotions into the analysis, you will get a global view of the emotional journey of your customers. And more generally, what emotions they feel and express. You will be able to compare between years, surveys, questions, customer segments, etc. 

You will be able to build your own benchmarks and see periodic evolutions on your survey results, in parallel to CSAT or quantitative scores. 

However, you may be wondering how emotions allow for a more refined analysis than sentiment analysis?

To answer this question as simply as possible, we must ask ourselves the question of prioritizing irritants. Indeed, after having performed a sentiment analysis, how can you know, without reading the verbatims, which negative comment is more important than the others?

Where sentiment analysis will segment the textual data into three distinct sentiments (positive, negative or neutral), emotional analysis allows to segment the comments into 6 so-called primary emotions (joy, positive surprise, fear, sadness, anger and disgust).

These emotions are said to be primary because they are common to all human beings regardless of age, gender, origin, etc.

If we go back to our question of prioritizing irritants, emotional analysis will allow us to go much further by detecting 4 negative emotions. And this is important because depending on the emotion detected, the action to be taken will not be the same at all.

Indeed, it is easy to understand that a customer who expresses fear will not need the same response as an angry customer. Without this emotional layer, with only an analysis of the feelings, it will be extremely difficult to correctly answer the customer's expectations. 

Imagine a customer expressing disgust with a product. If you offer him a discount to buy this product, it will not work, for 2 main reasons:

• because the customer is already looking for a complete change
• because this emotion is often uncontrollable, stronger than their mind and will. 

Offering the same answer over and over again is like saying: "I don't understand what you're telling me". 

Here is a table that details the different action plans expected by clients depending on the emotion expressed:

As you can see, emotional analysis allows you to go much further and understand in-depth the real experience of your customers. It is even more effective if you conduct surveys with open-ended questions.

Moreover, if you already use satisfaction indicators such as NPS (Net Promoter Score) or CSAT (customer satisfaction), then combining these indicators with emotional analysis will give you even more convincing results.

Combining satisfaction indicators and emotional analysis

Combining traditional satisfaction indicators with emotions would help you leverage the simplicity of average CSAT scores on global questions or a few strategic questions AND the depth and added value of customer feedback. 

By coupling the two approaches, you will be able to better understand the evolution of satisfaction over time and compare results over different periods... 

In the example below on a banking case, we can see that there has been a decrease in
satisfaction of 0.7 points. 

This is mainly linked in the comments to the "Attention to customer", “Fees”, “Respect to customers” topics whose services has clearly deteriorated, and which explains 0.6 points of decline in satisfaction over the year. 

For some topics, you can go beyond satisfaction monitoring to manage customer emotions
throughout the customer journey(s). 

And even if you can't perform a robust analysis, starting to collect open-ended feedback will help you get historical data, which is also valuable for assessing trends and manually checking for weak signals. 

As you can see, each type of analysis has its strong point. While sentiment analysis is more suitable for social listening and social networks, emotional analysis allows you to fully exploit survey verbatims and online opinions. 

The solution we offer at Q°emotion is a SaaS platform for automatic emotional and semantic analysis of customer feedback. The tool allows the analysis of comments from all sources and in more than 30 languages. If you want to know more, you can discover a case study below or book a live demo by clicking here.