Amazon Comprehend Conditions
An Amazon Comprehend condition matches PII entities that Amazon Comprehend detects in a column’s sampled values (a sampled subset; see Sampling), joining Google DLP and Snowflake Native as the classifier rule builder’s external, ML-based signals.
Amazon Comprehend detects PII entities in English text.
The ALTR Managed - AWS Comprehend default collection provides ready-made classifiers for a selection of entity types. See Third-Party Default Collections.
Fields
Section titled “Fields”-
Entity type(s) – a searchable multi-select of Amazon Comprehend PII entity types. For the entity types Amazon Comprehend can detect, see AWS’s Detecting PII entities reference. Selecting multiple entity types combines them with OR logic: the condition passes if any selected type is detected at or above the Match Threshold.
-
Match Threshold – a score gate from 0-1 (not a percentage); a detection below it doesn’t satisfy the condition. Defaults to
0.5.
Example
Section titled “Example”This condition matches columns Amazon Comprehend identifies as an email address, at a minimum score of 0.8:
{ "target": "AMAZON_COMPREHEND", "entity_type": "EMAIL", "minimum_score": 0.8 }Behavior
Section titled “Behavior”In Match Confidence scoring, an Amazon Comprehend match is outranked only by Google DLP and Snowflake Native matches, and outranks a Regex, Data Entropy, Column Content, Column Name, Data Location, Data Length, or Column Size match on the same column.
Availability
Section titled “Availability”Amazon Comprehend evaluates on Snowflake ALTR Native jobs using ALTR Hosted processing (not In-Warehouse processing) and on OLTP ALTR Native jobs. On processing paths where it can’t run, the condition is treated as not evaluated rather than failing the job. See Classification Processing Location for the full picture.
Amazon Comprehend is opt-in per job: it only runs when all three hold – a classifier in the job includes an Amazon Comprehend condition, the job’s selected condition types include Amazon Comprehend, and the job samples row data.
On OLTP, Amazon Comprehend runs through your own AWS account rather than ALTR’s. For a file classification job, the OLTP Classification Agent sends the file’s text to Amazon Comprehend as a document, in overlapping chunks for large files. See the OLTP Classification Agent for how ALTR handles credentials for this and other external conditions.