AWS Comprehend

Submits up to 5,000 bytes of text and returns the ISO 639-1 language code with a confidence score, letting your agent route multilingual content without any manual triage.

Scores a text passage as Positive, Negative, Neutral, or Mixed with per-label confidence values, giving your agent a precise read on tone before it decides what to do next.

Extracts named entities — people, organisations, locations, dates, quantities, and more — from a text passage so your agent can act on structured facts hidden in unstructured copy.

Processes up to 25 text documents in a single API call for entity extraction, letting your agent handle bulk NLP jobs without looping one document at a time.

Returns the most salient noun phrases from a text passage, giving your agent a concise topic fingerprint without requiring a custom ML model.

Extracts key phrases from up to 25 documents in one request, making large-corpus topic analysis practical without repeated single-document calls.

Analyses sentiment across up to 25 text documents simultaneously, enabling bulk emotional-tone assessment across survey responses, reviews, or messages in one API call.

Parses a text passage into tokens annotated with parts of speech — noun, verb, adjective, and more — giving your agent structural linguistic insight for grammar-aware automation.

Identifies and labels personally identifiable information — names, addresses, phone numbers, SSNs, credit card numbers, and more — so your agent can redact or flag sensitive data before it spreads.

Returns a binary flag indicating whether a text passage contains any PII, letting your agent make a fast gate decision without the overhead of full entity extraction.

Kicks off an asynchronous Amazon Comprehend job that clusters a large document collection into latent topics using ML, returning a job ID your agent can poll or react to via EventBridge.

Launches an asynchronous batch sentiment analysis job for large document sets stored in S3, enabling high-volume emotional-tone scoring that would time out as synchronous calls.

Submits a large S3 document corpus to an asynchronous entity extraction job, letting your agent process millions of words of text without synchronous API limits.

Runs a text passage through a trained Amazon Comprehend custom classifier endpoint and returns predicted class labels with confidence scores, enabling domain-specific categorisation without rule writing.

Launches an asynchronous job that scores sentiment toward specific named entities in a document corpus, giving your agent entity-level tone analysis across large text batches.

Retrieves the current status, configuration, and output location for a running or completed asynchronous sentiment detection job, letting your agent check progress without polling CloudWatch.

Fires via EventBridge when an asynchronous sentiment detection job reaches COMPLETED status, letting your agent immediately pull and process the results without polling.

Fires via EventBridge when an asynchronous sentiment detection job reaches FAILED status, letting your agent alert the team and optionally retry before the error is noticed manually.

Fires via EventBridge when an asynchronous entities detection job finishes successfully, enabling your agent to trigger downstream indexing or enrichment steps without any polling loop.

Fires via EventBridge when an asynchronous topic modelling job over a document corpus finishes, letting your agent immediately ingest and distribute the topic cluster results.

Fires via EventBridge when an asynchronous PII detection job finishes, enabling your agent to trigger redaction, quarantine, or compliance reporting steps immediately on the results.

Fires via EventBridge when a custom classifier endpoint transitions state — such as IN_SERVICE, UPDATING, or DELETING — so your agent can pause dependent workflows or notify the ML team automatically.