Sufficient consensus
Sentiment analysis classifies text as positive, negative or neutral, with aspect-based variants that attach sentiment to specific features. It is applied to reviews, support tickets and social posts.
At scale it reliably detects direction and change.
What the result pages leave out
These belong to the subject. People ask about them. They are missing from the agreed coverage.
- The specific complaint. A negative score is actionable only after someone reads the sentence that produced it.
- Silence from the disappointed. Most dissatisfied people say nothing at all, so the corpus over-represents the vocal.
- Neutral is the largest and least examined class. Factual descriptions of a failure are frequently classified as neutral.
- Irony and understatement. Both are common in exactly the cultures and industries where the stakes are highest.
- Satisfaction that did not last. Sentiment is captured at one moment. The verdict that matters arrives months later.
What people actually want to know
- What exactly are they unhappy about?
- Who is not writing to us at all?
- What is in the neutral pile?
- Do the people who praised us six months ago still use the product?
Read fifty of the actual sentences. The chart tells you that something happened. The sentences tell you what.
More in computational linguistics
Extractive summarizationThe summary is now the article for most of the audience.Keyphrase extractionIf your real subject is named once, it is not your subject as far as the system is concerned.Semantic searchThe engine got better at knowing what you meant.Named entity recognitionBeing an entity is a prerequisite for being understood.Topic modelingThe strongest finding in any corpus analysis is usually the subject that is missing.Question answeringA confident single answer is the format.Word sense disambiguationYour page competes in whichever meaning the system assigned to it.