NLP Pro
Key phrases, aspect sentiment, custom classification (zero-shot), entity linking, and conversational PII.
The capabilities that take Azure AI Language and AWS Comprehend from "basic" to "useful" — at a fraction of the price.
What you get
/v1/nlp/keyphrasesKey Phrases (YAKE-style)
Top-N ranked phrases from text via statistical scoring (TF + position + casing + stopword filter). No ML model, predictable cost, blazing fast.
Engine: Brainiall Key Phrases engine
/v1/nlp/sentiment/aspectAspect Sentiment
Per-aspect sentiment scoring. Send a product review + aspects like [camera, battery, screen, price] and get a per-aspect verdict with mention spans.
Engine: Brainiall Aspect Sentiment engine
/v1/nlp/classify/customCustom Classification (zero-shot)
Define your labels at call time — no training data, no model upload. Returns predicted label + per-label scores. Single or multi-label.
Engine: Brainiall Custom Classifier engine
/v1/nlp/entities/linkEntity Linking (Wikidata)
Detect entities AND link each one to its canonical Wikidata Q-id, label, and description. Disambiguate "Apple" the company from "apple" the fruit.
Engine: Brainiall Entity Linker engine
/v1/nlp/pii/conversationConversational PII (multi-turn)
Detect PII across a multi-turn dialogue with cross-turn coreference. Same email, name, or ID gets the same stable entity_id no matter how often it appears.
Engine: Brainiall Conversational PII engine
How we compare
| Provider | Equivalent surface |
|---|---|
| AWS Comprehend | Has key phrases, targeted sentiment, custom classification (requires training), no entity linking, conversational PII not native |
| Azure AI Language | Has all five plus more, but charged per-record per-feature, custom classification requires training jobs |
| Brainiall | All five bundled, custom classification is zero-shot (no training), entity linking included. NLP-Suite priced per 1k records. |
Pricing
NLP Pro endpoints share your existing Brainiall NLP, Document, and Speech AI usage — no separate bundle subscription. The Free tier covers ~100-1000 calls/month per endpoint.