Semantic and Sentiment Analysis: What Are the Best Sources of Consumers’ Thoughts?

Semantic and Sentiment Analysis: What Are the Best Sources of Consumers’ Thoughts?

What do your users really feel and think about your product, service, or content?

Datafication is something we always talk about. We have data flows, we want as much data as possible, and we try to find a way to interpret it. Data tells us a lot but not about the feelings of users. Imagine you have a user who has a subscription to your service. They spend time and money on your platform, consuming your content. You have as much data about them as possible, yet once this user cancels their subscription and starts using another platform, you try to win them back—sending offers, emails—but now, they are classified as churned. However, you still don’t know why this user left and what you actually need to offer to bring them back.

It sounds like we are discussing not just a relationship with a user but a relationship with a significant other. But come on, aren’t your users significant to you? Of course, they are. And what is the best way to prevent a relationship catastrophe? Listening.

Direct questions you ask your users do not always work well. A user must be extremely loyal and have plenty of free time to fill out occasional surveys about user satisfaction. If the user is no longer satisfied and decides to switch to another service, your final attempt with a five-question poll asking why they are leaving is a last-ditch foul, but you should have listened earlier.

We can reflect on this for a long time, but two key methods really help you listen to your users and extract as many insights as possible: the first is sentiment analysis, and the second is semantic analysis.

Sentiment Analysis

Sentiment analysis encodes the emotional tone of a text. Usually provided by NLP, this type of analysis can be applied to any text: social media posts, threads, comments on YouTube or Twitch, survey results. This analysis is less about deep insights and more about understanding users’ feelings and satisfaction with a product or content. Typically, sentiment analysis categorises data into three groups: positive, negative, and neutral. All comments, responses, and messages—depending on your goal and the platform for analysis—fall into these categories.

For example, if you release a trailer for a game or film, a demo of a new feature, or plan a rebranding for the product, you can gauge the general impression of your current or potential consumers. How do they feel about it? Good, bad, or indifferent? This allows you to address issues if something was not well received.

Semantic Analysis

Semantic analysis provides specific answers. If sentiment analysis reveals that users did not like your trailer or feature, you need to find out why. Of course, you can try to survey a sample of people who had a negative reaction, but a sample may not capture all insights. More importantly, consumers, when asked directly, are not as honest as they are when they simply interact. When faced with someone responsible for the product, their behaviour changes—they become less critical, more polite, and ultimately less insightful.

Semantic analysis helps by categorising discussions based on topics relevant to the content users are engaging with. For example, if it’s a film or game trailer, you can determine which aspects users discussed the most: graphics and CGI? Voice acting? Facial animations? An unrealistic plot? Predictable gameplay? Or perhaps the controversy surrounding the producer? Sentiment analysis may show that 60% of comments were negative, but the underlying reason for this negativity may vary widely. A controversy involving a key figure in the project is a very different issue from an unpleasant voiceover.

Combining Sentiment and Semantic Analysis

It’s best to combine both analyses. First, use sentiment analysis to understand the general attitude toward your content—was it positive, negative, or neutral? Once you have an answer, apply semantic analysis. It identifies the most and least discussed topics as well as the words and concepts users mention most often. Such a simple technique can provide valuable insights.

If you want to dig even deeper, use sentiment analysis again, but this time, apply it to specific sets of messages or comments. For example, suppose semantic analysis shows that a significant portion of comments is about CGI. Are most of these comments positive or negative? Another popular discussion topic is the production team—what is the sentiment there? Good, bad, or neutral?

It’s not just about raw data—it’s about interpretation, not subjective but direct and objective. You may have 10,000 comments under posts about two games you promoted. Both posts have excellent engagement rates, suggesting your brand is popular. However, sentiment analysis reveals that the first post’s 10,000 comments are mostly negative, while the second’s are predominantly positive. Meanwhile, semantic analysis shows that the positive comments were about the winning team, whose fans flooded the post to celebrate, whereas the negative comments stemmed from technical issues on your platform. Despite high engagement rates, analysis reveals the actual reasons behind the activity.

Finding the Right Data Sources

Of course, data is necessary for such analysis, and social media is an obvious feedback source. But is it truly the best? Are those commenting on your posts actual users, or did they simply see the post in their feed, leave a comment, and move on?

For some businesses, this distinction is not critical. If you are marketing a new product that is available everywhere, social media is ideal. However, if you are marketing a service, platform, or content exclusive to a specific platform, analysing social media may not provide the insights needed to refine your strategy or improve your product. In such cases, in-app messaging is the most valuable data source. Users share their emotions and opinions while consuming content, not through structured surveys but in organic discussions with each other. Such feedback is far more sincere and, therefore, more insightful.

How to Gather In-App Data

Public Live Chats

When users have a space to openly communicate about topics they enjoy, they become honest. They discuss both the advantages and disadvantages of your platform and may even debate among themselves. This not only boosts retention and engagement but also provides you with valuable, unsolicited feedback—at no additional cost. Public chats can be integrated in various ways, such as a general lobby for all conversations or dedicated chats for specific events or content pages.

Comment Sections

Like event chats but designed for long-term discussions rather than real-time interaction, comment sections provide direct feedback at all times. Media platforms, streaming services, and marketplaces can all introduce comment sections to continuously receive user insights. Publishers claim that comment sections also increase time spent on the platform and subscriber numbers, as users often need to sign in to participate.

Moreover, an essential aspect of in-app data for sentiment and semantic analysis is that you own the data. On social media, inappropriate comments may not be visible, meaning they are excluded from analytical tools. However, on your platform, even hidden comments can be analysed, allowing you to understand what truly frustrates your users.

In-App Live Streaming

Live streaming is an exceptional source of insights because it is spontaneous, engaging, and real-time. Users are accustomed to following Twitch influencers who analyse content or products, and they trust these influencers. If an influencer streams directly on your platform, users will still be direct and honest.

How Semantic Analysis Works on Watchers Platform

Our clients can analyse messages sent to any chat using AI-powered tools. Messages from a given period are forwarded to the tool, and clients can select from various analysis prompts or customise them for optimal results. The system categorises messages, allowing clients to understand which topics users discussed most, the sentiment behind them, and the most frequently used words and phrases.

Semantic analysis can also serve specific purposes. For instance, one recent analysis by one of our clients revealed that their public chat contained messages in six languages. However, 60% of the messages were in French, with the most common words being gagner, real, psg, lyon, and mbappe. With this insight, they knew which events to prioritise on their platform’s homepage.

***

Sources

- YouTube Comment Analyzer Using Sentimental Analysis

- Brand Meaning and Its Social Categories: A Semiotic Approach for Future Marketing

- Semiotics in Marketing

Get in Touch

If you want to partner with us or get more details, please schedule a demo meeting by filling in the form. We'll get in touch with you shortly.