Sentiment analysis
Sentiment analysis API with your own scale
Short answer
A sentiment analysis API scores how positive or negative a piece of text is. With a decision model you define the scale yourself, such as five levels from very negative to very positive, and get the most likely level, a weighted score and a probability per level. You can also ask about specific aspects like price or support.
Classic sentiment tools return positive, negative or neutral. That's rarely what a business needs. Support teams want to know who is frustrated enough to churn, product teams want sentiment per feature, and marketing wants to separate mild praise from enthusiasm.
Defining your own scale and aspects turns sentiment from a vanity metric into something you can route and act on.
Get Early Access to ClassifierHub: 2× credits in your first paid month.
Scores on your scale
Use a score question with the levels you care about. The response includes the winning level, a probability-weighted raw score (useful for averages and trends) and a probability per level.
curl https://classifierhub.com/v1/score \
-H "Authorization: Bearer $CLASSIFIERHUB_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"input": "The app is fine I guess, but support took four days to answer.",
"instructions": "How does the customer feel about the company overall?",
"levels": ["Very negative", "Negative", "Neutral", "Positive", "Very positive"]
}'
# -> { "score": 1, "raw_score": 1.4, "label": "Negative", "confidence": 0.61,
# "probabilities": [0.05, 0.61, 0.3, 0.04, 0], "usage": { "credits": 1, ... } }The ClassifierHub API opens to Early Access members first.
Aspect-based sentiment
Overall sentiment hides the details. "Love the product, hate the delivery" is neutral overall but very useful per aspect. Ask one score per aspect you track (product, price, delivery, support), plus a yes/no for whether each aspect is mentioned at all. Up to five outputs about the same input cost 1 credit together.
Acting on sentiment
- Escalate tickets where sentiment is very negative and the customer is on a high-value plan.
- Ask happy reviewers for a public review; route unhappy ones to support first.
- Track the average raw score per week and per aspect to see whether changes land.
Frequently asked questions
Related guides
Last updated . ClassifierHub is an independent product built on top of the Jev decision model, accessed through OpenRouter. It is not affiliated with or endorsed by TypeSafe or OpenRouter.