Sentiment analysis explained without the jargon
Sentiment analysis is software reading a piece of text and judging whether it is positive, negative or neutral about a subject. Simple to say, historically hard to do.
The old way: word lists
Early tools counted good and bad words. "Great" scored plus one, "terrible" minus one. This falls over immediately on real language: "great, another price rise" contains a positive word and a wholly negative meaning. British English, rich in understatement and sarcasm, breaks word-list systems roughly constantly.
The new way: language models
Modern AI models read the whole sentence in context, the way a person does. They handle "not bad at all" (positive), "well that's just brilliant isn't it" (negative), and "the app is awful but the staff sorted it instantly" (mixed, leaning positive about service). Accuracy on everyday text has moved from coin-flip territory to broadly human-level.
What still goes wrong
Sentiment toward whom is the hard part. "They gave that lot a sponsorship deal? Fools." is negative, but about which party? Good systems are told who the client is and score sentiment toward them specifically. In-jokes, niche slang and heavy irony still produce occasional misreads, which is why serious setups keep a human eye on negative classifications.
What to do with it
One mention's score matters little. The trend is the product: net sentiment this month versus last, and which topics drive the negative side. That is a management report, not a technology demo.
MyView watches every mention of your brand and tells you what needs a reply.
Book a demo