Amazon Web Services → Microsoft Azure
Azure equivalent of Amazon Comprehend
The Azure equivalent of Amazon Comprehend is Azure Language.
About Amazon Comprehend
| Provider | Amazon Web Services |
|---|---|
| Status | Generally available Topic modeling, event detection, and prompt safety classification are no longer available to new customers (maintenance, Mar 31, 2026). Other features are unaffected. |
| What it is | NLP. Entities, key phrases, sentiment and targeted sentiment, language detection, PII, syntax, toxicity, and custom classification/entity recognition. |
| Documentation | docs.aws.amazon.com |
| Pricing | Official pricing page |
Azure mappings (4)
| Capability | Azure equivalent | Match |
|---|---|---|
| Image & text content moderation | Azure AI Content Safety | Direct match |
| Text analytics (entities, sentiment, key phrases) | Azure Language | Direct match |
| Custom text classification & entity recognition | Azure Language | Direct match |
| PII detection & redaction in text | Azure Language, Foundry guardrails (content filters) | Direct match |
Vision & documents(1)
Image & text content moderation Direct match
Rekognition content moderation (and Comprehend toxicity detection for text) maps to Azure AI Content Safety.
- Rekognition Batch Image Content Moderation entered maintenance on Mar 31, 2026. Real-time moderation APIs are unaffected.
- Azure Content Safety returns multi-severity scores for hate, sexual, violence, and self-harm, and supports custom categories (preview) and blocklists. The old Azure Content Moderator is retired.
Language & translation(3)
Text analytics (entities, sentiment, key phrases) Direct match
The Azure equivalent of Amazon Comprehend is Azure Language. Note that Azure now labels sentiment, key phrases, summarization, and entity linking as "legacy" capabilities, supported for existing implementations.
- Azure's "core" (recommended) capabilities are PII, language detection, NER, custom NER, and Text Analytics for health. Sentiment and opinion mining, key phrases, entity linking, summarization, and custom text classification are "legacy".
- Comprehend topic modeling, event detection, and prompt safety classification are closed to new customers (Mar 31, 2026). AWS points to Bedrock LLMs and Guardrails instead.
- Both vendors steer new open-ended NLP (topics, events, summaries) toward LLMs (Bedrock or Foundry models).
Custom text classification & entity recognition Direct match
Comprehend custom classification and custom entity recognition map to Azure Language custom text classification (legacy) and custom NER (core).
- Comprehend custom models accept plain text, and also PDF, Word, and images (for custom classification and entity recognition).
- On Azure, custom NER is a "core" capability, while custom text classification is "legacy". For new classification work, Microsoft leans toward Foundry models.
PII detection & redaction in text Direct match
Comprehend PII detection maps to Azure Language PII detection, which is a "core" Azure Language capability covering text, conversations, and documents.
- Azure also offers Document PII, a playground in Foundry, and synthetic-replacement anonymization (preview).
- For GenAI traffic, PII masking is built into Bedrock Guardrails (sensitive information filters) and Foundry guardrails (PII risk, preview).