Why Sentiment Matters

Sentiment shapes many of the decisions we make every day. Whether it’s reading product reviews, browsing social media comments, or listening to someone’s feedback, we rely on the opinions of others to guide us. But why do these feelings and opinions matter so much, and how can we understand them better?

What is Sentiment, Really?

Sentiment analysis is a systematic process for evaluating the expressed emotions within the opinions shared by people. As a field of study within computer science, sentiment analysis didn’t really take off until the early 2000s. Thanks to the internet, and how easy it is for us to leave reviews and opinions online, the need to analyze this rich data became essential. Now, sentiment analysis helps us gauge how people feel about products, services, or even entire brands.

Sentiment, at its core, is about feeling. The word itself traces back to Latin and French roots, originally meaning something personal, an emotional response. By the 18th century, people began using phrases like, “My sentiments exactly,” to express alignment with others’ feelings. Today, when we talk about sentiment, especially in the context of technology, we’re referring to understanding collective opinions - whether they’re positive, negative, or somewhere in between. We're known for making our decisions before digesting all the factual information available.

“We are not thinking machines that feel; rather, we are feeling machines that think.”

– Antonio Damasio, neuroscientist and author of Descartes’ Error

Cue the need for sentiment analysis to help us analyze the opinions expressed by others.

Why Sentiment Matters to Us All

We are social beings by nature, hardwired to care about what others think. This is why, before making a purchase, 82% of us read reviews to understand how others feel about something we’re considering buying. We trust those reviews because they offer us a peek into what others experienced, helping us decide what’s worth our time, money, or energy. Thanks to the internet, this pool of shared information and experiences runs deeper than ever before.

This deeply ingrained reliance on shared experiences and opinions isn’t just a modern phenomenon; it reflects a fundamental aspect of human nature that some scientists call “collective intelligence.” Where we once turned to family, friends, or community members for advice, we now include others’ opinions online to provide us with extra understanding and context. This shared knowledge reduces our risk and helps us make decisions. Interestingly, our reliance on collective intelligence is so significant that some scientists attribute it to a reduction in our brain size over time. The skulls of men and women today are on average 12.7% smaller than that of Homo sapiens who lived during the last ice age. Our species has evolved to use shared experiences to guide us, whether through personal interactions or online reviews - and this collective intelligence is telling the story about your product or brand on some corner of the internet!

One fascinating aspect of sentiment is how quickly we pick up on it. Emotional content in reviews - whether it’s excitement, frustration, or disappointment - jumps out at us. Studies show that emotional reviews are easier for us to digest and process. Positive, emotionally charged reviews make us more likely to feel confident that we're about to make a good choice. Negative reviews, on the other hand, make us stop and think harder, prompting us to consider the potential problems associated with the product. Mixed reviews, have been shown to be interpreted as most helpful, thanks to the blend of positive and negative feedback which to us, feels balanced and credible.

We might still make a purchase even if there are some negative opinions about a product, as long as the reasons for the negativity aren’t deal breakers. Studies on review sentiment often reveal seemingly conflicting outcomes when it comes to purchasing decisions. A product that received some negative reviews does not automatically mean people will avoid buying it altogether. If a product meets a need, and some reviewers mention its limitations, a buyer might still proceed with the purchase - armed with additional insight from the reviews left by others.

For companies, this collective reliance on reviews and sentiment means you are unable to shape the brand story on your own. Understanding the online narrative or story as told by your customers, can help you be more strategic. Often times focus groups, or usability studies, shape insight from a certain angle. But what about the louder voices online? These need to be factored into your research road map too. Understand which features or products make your customers happy and let this guide you to keep providing this. Understand where frustrations are bubbling up? Fix those. Sentiment analysis tools help bring quantifiable data from the treasure trove of free-text data about any topic. Good tools reveal common themes and paint a clear picture of what people love, and what they don't. Which brings us to the crux - how to get beyond just positive/negative!

Getting beyond simple positive vs negative

Basic sentiment analysis tools label opinions as positive, negative, or neutral. Some even include a mixed category where both positive and negative emotions are detected. But modern tools like AddMaple can go so much further. They can look at the specific things people mention in reviews, and can analyze these for sentiment. Some call this topics or aspect analysis. Read on to see how AddMaple helps you take this even further.

✅ High level themes by Sentiment:
Instead of detecting and classifying a review as a whole as simply “positive” or “negative,” tools like AddMaple can pinpoint exactly which parts of the review, relating to specific aspects of the product or service, are linked to positive, negative, or mixed sentiment. For example, themes such as a product’s design, its durability, or the quality of customer service can be analyzed to determine their individual sentiment.

Many tools, such as AWS Comprehend or Google Natural Language, rely on predefined themes that they attempt to apply to reviews, often resulting in miscategorized data. For instance, you might see ‘Location’ as a label applied to a review mentioning the kitchen, or find most aspects of the review labeled as ‘Uncategorized.’

Because AddMaple uses a combination of approaches, it can support nuanced analysis of reviews, whether they are about restaurants, open-ended feedback for St John’s Ambulance, or other industries. A relatable example is restaurant reviews: while a restaurant may find that the sentiment about its food is overwhelmingly positive, it might also discover that service has mixed sentiment, with concerning levels of negative feedback. Legacy tools that only provide an overall classification of positive or negative miss out on these critical distinctions.

✅ Next Level - Prompting for topics of interest and analyzing those by Sentiment:
AddMaple is one of the only tools that takes topic sentiment analysis to the next level. With prompting, you can ask AddMaple to analyze reviews while focusing on specific areas of interest. For example, a restaurant owner could extract mentions of specific foods from reviews, analyze those excerpts for sentiment, and generate a clear breakdown of positive, negative, and mixed sentiment for each food item.

This means you can see that your potatoes, for instance, are linked with negative sentiment more than any other dish. AddMaple doesn’t just analyze features broadly; it delves into the finer details of those features, providing unparalleled insights into what drives sentiment. The output, a Likert chart, shows you the ranking of each food by its sentiment proportions. This is truly best in class. No other tools do this!

Why we built and released Sentiment Lab?

We built Sentiment Lab to give everyone the opportunity to explore how 12 different sentiment models perform on various datasets, including your own text. It allows you to see which models grasp nuance and sarcasm, and which come closest to human-like sentiment evaluation. We enjoyed the process of comparing these models so much that we decided to release the laboratory for you to experiment with, letting you see which models align most with your own text and sentiment perspective.

Ultimately, sentiment matters because it’s about people - understanding how they feel. By paying attention to sentiment in reviews, social media, or feedback forms, we gain deeper insights into the collective opinion online which influences people daily. This becomes even more powerful when text is analyzed into thematic extracts, sub-themes, and other aspects by sentiment, helping us see which areas are more positive or negative relative to each other. These narratives shape our understanding and decisions and we are here to help you understand them better, more clearly, with the ability to ask specific questions. This is the next level of analysis we’re excited to bring to you with AddMaple.