Sentiment & context, explained simply
Sentiment analysis means guessing the feeling behind some text: positive, negative or neutral. Companies use it to sort thousands of reviews or messages quickly.
A simple word-counting model can be fooled easily. "Not bad at all" contains "bad" but is a compliment. "Oh great, it rained again" contains "great" but is a complaint.
People use context, tone and shared knowledge to catch these. Simple models miss them, and even advanced ones sometimes do. That is a good reason to check before trusting an AI’s reading of someone’s feelings.
Example: Tricky sentences
- "I love this!" → positive (easy)
- "This is not good." → negative, but a word counter may see "good" and say positive.
- "Brilliant, my phone died again." → sarcasm, negative. Hard for simple models.
Try this
- Write three sentences designed to fool a word-counting sentiment model.
- Say the same sentence in a happy voice and a grumpy voice. What information does text lose?
- Sort ten emoji into positive, negative and "it depends".
