Glossary
Opinion Mining
Extracting structured opinion data from unstructured text at scale.
Definition
Opinion mining is a field of natural language processing focused on extracting structured opinion data from unstructured text. It encompasses sentiment analysis, stance classification, aspect-based opinion extraction, and narrative identification — any technique that converts raw text into an explicit representation of who thinks what about what.
How it works
- 1
Text is collected from comments, reviews, forums, or surveys.
- 2
Entities (brands, people, topics) and attributes (features, policies) are identified.
- 3
Opinions about each entity-attribute pair are classified for polarity and stance.
- 4
Results are aggregated into structured datasets for analysis or reporting.
Why it matters
Unstructured text contains enormous amounts of actionable intelligence that is invisible without analysis. Opinion mining makes that intelligence legible — turning thousands of free-text comments into structured breakdowns of who holds which view, how strongly, and with what confidence.
Related distinctions
Opinion mining vs sentiment analysis
Sentiment analysis is a subset of opinion mining focused specifically on emotional polarity. Opinion mining also covers stance, aspect-based opinions (which specific features are liked or disliked), and argument mining.
Opinion mining vs text analytics
Text analytics covers all forms of quantitative text analysis including classification, extraction, and summarisation. Opinion mining is the subset of text analytics specifically concerned with extracting opinion, position, and subjective viewpoint.
Frequently asked questions
What is opinion mining?
Opinion mining is the automated extraction of structured opinion data from unstructured text. It includes sentiment analysis, stance classification, and aspect-based analysis — any technique that converts raw text into an explicit record of who thinks what about what.
What is the relationship between opinion mining and sentiment analysis?
Sentiment analysis is one technique within opinion mining. It classifies emotional polarity. Opinion mining is the broader field that also includes stance detection, aspect extraction, argument mining, and opinion summarisation.
What are the main applications of opinion mining?
Key applications include: product review analysis (which features customers like or dislike), political opinion research (how public sentiment on policies shifts), brand monitoring (reputation tracking over time), and customer feedback triage (automatically routing complaints by category).
How accurate is opinion mining?
Accuracy varies significantly by technique and domain. Sentiment analysis on formal text can reach 90%+ accuracy. Stance classification on short social media comments typically achieves 75–85%. Aspect extraction and sarcasm detection are harder, typically 65–80%.
See opinion mining in practice
Narativ applies stance analysis, narrative clustering, and engagement weighting to live comment sections — from £1 per post.