Glossary
Sentiment Analysis
Emotional tone in text — widely used, but limited for opinion research.
Definition
Sentiment analysis classifies the emotional tone of a piece of text as positive, negative, or neutral. It is the most widely deployed form of natural language processing in social listening tools, but it measures how something is said rather than what position is being taken.
How it works
- 1
Text is processed by a language model or lexicon-based classifier.
- 2
Each piece of text receives a polarity score: positive, negative, or neutral.
- 3
Scores can be aggregated at post, campaign, or brand level.
- 4
Trends over time are tracked to measure shifts in public mood.
Why it matters
Sentiment analysis is a quick, scalable proxy for audience mood. It works well for brand monitoring and customer feedback categorisation. Its limitation is that polarity does not reliably indicate whether someone supports or opposes a position — a critical gap for PR, public affairs, and policy research.
Related distinctions
Sentiment analysis vs stance analysis
Sentiment measures emotional tone. Stance measures directional alignment. For campaigns that want to know "are people for or against this?", stance analysis produces a more accurate and actionable answer.
Sentiment analysis vs opinion mining
Opinion mining is the broader field that includes sentiment as one technique. Opinion mining also covers aspect-based analysis (which features are liked/disliked), stance classification, and narrative extraction.
Frequently asked questions
What is sentiment analysis?
Sentiment analysis is a natural language processing technique that classifies the emotional tone of text as positive, negative, or neutral. It is widely used in social listening, customer feedback tools, and brand monitoring.
What are the limitations of sentiment analysis?
Sentiment analysis measures how text sounds emotionally — it does not reliably tell you what position the author holds. Sarcasm, irony, and hedged language frequently cause misclassification. For opinion research, stance analysis provides a more accurate signal.
When should I use sentiment analysis vs stance analysis?
Use sentiment analysis for broad mood tracking and customer feedback categorisation. Use stance analysis when you need to know whether an audience actually supports or opposes a specific claim — for example, in a product launch, crisis, or public affairs context.
Is sentiment analysis accurate?
Modern transformer-based sentiment models achieve 80–90% accuracy on benchmark datasets. Real-world accuracy is lower due to domain-specific language, sarcasm, and short social media text. Accuracy improves significantly when models are fine-tuned on domain-specific data.
See sentiment analysis in practice
Narativ applies stance analysis, narrative clustering, and engagement weighting to live comment sections — from £1 per post.