Sentiment Analysis

Sentiment analysis is the use of AI to determine the emotional tone of text. It classifies text as positive, negative, or neutral — and more sophisticated systems detect specific emotions like frustration, excitement, sarcasm, or urgency.

How it works

Modern sentiment analysis uses language models that have learned the relationship between words and emotional tone from millions of examples. The sentence "This product exceeded all my expectations" is classified as positive. "I have been waiting three weeks and still no response" is classified as negative.

Simple sentiment systems use keyword matching — counting positive and negative words. More advanced systems understand context. "This is not bad" is positive despite containing the word "bad." "I could not be less impressed" is negative despite containing "impressed." Modern LLMs handle these nuances naturally.

Common applications

• Customer feedback analysis: Automatically categorise thousands of reviews, support tickets, or survey responses by sentiment
• Social media monitoring: Track public sentiment about your brand, product, or industry in real time
• Voice of customer: Identify emerging complaints or praise patterns before they become trends
• Employee experience: Analyse internal survey responses and communication patterns
• Market research: Gauge public reaction to product launches, announcements, or competitor moves

Beyond positive/negative

Advanced sentiment analysis goes further:

• Aspect-based sentiment: Identifies sentiment about specific features ("The camera is excellent but the battery life is poor")
• Emotion detection: Classifies specific emotions (anger, joy, fear, surprise)
• Intent detection: Determines whether a message requires action (complaint, question, compliment)

Using LLMs for sentiment analysis

With modern LLMs, you do not need a dedicated sentiment model. You can simply prompt Claude or ChatGPT: "Classify the sentiment of each of these customer reviews as positive, negative, or neutral, and identify the main topic." This flexibility means sentiment analysis is now accessible to anyone who can write a prompt.

Accuracy considerations

Sentiment analysis struggles with sarcasm, cultural context, domain-specific language, and mixed sentiment (a review that praises some aspects and criticises others). For business-critical applications, always validate AI sentiment analysis against human judgement on a representative sample.