Route daily requests into the existing returns, delivery, and faults categories. In a separate monthly analysis, explore an unlabeled collection to find similar concerns that the current taxonomy may miss.
Compare assigning known categories with discovering groups in a collection. Automatically approving and sending replies is outside these analysis requirements.
Choose TWO appropriate explanations of the analysis goals.
Distinguish classification into known categories from discovery of groups in unlabeled text. Interpret clusters rather than equating them with business labels, exact string matching, or generated and sent replies.
Detailed explanation
Group discovery is separate from text generation and external sending. An analysis output does not establish completion of unrequested approval or sending actions.
Classification need not be exact category-name matching. Models may recognize paraphrases, but performance on actual examples still needs evaluation.
Similarity groups do not necessarily match predefined business categories. Inspect their contents to assign names and operational meaning.
This assigns defined categories to messages. Check category definitions and corresponding examples.
This explores structure rather than predicting supplied answer classes. Whether discovered groups are useful business categories requires separate assessment.
Check official documentation
Compare the following points with the official documentation.
Compare Comprehend custom classification with AWS unsupervised-learning definitions. Use invented requests to contrast assigning predefined categories with exploring groups of similar text and interpreting their meaning.
Official references
What to confirm
Distinguish prediction of known categories from exploration of unknown structure. Do not assume cluster names are automatically correct business categories or that replies have been generated, approved, or sent.
Key points
- Known categories
- Unknown groups
- Group interpretation
- Classification and matching
Notes
- Environment: Current AIF-C01 Task 1.1 and related AWS primary-source review using invented examples; no model training or inference is performed.
- Compare Comprehend custom classification with AWS unsupervised-learning definitions. Use invented requests to contrast assigning predefined categories with exploring groups of similar text and interpreting their meaning.
- Distinguish prediction of known categories from exploration of unknown structure. Do not assume cluster names are automatically correct business categories or that replies have been generated, approved, or sent.
Foundation review
Separate known answers from unknown structure
Daily routing uses predefined categories; monthly exploration examines groups of similar concerns. The same type of input text can support different goals and outputs.
Interpret group meaning
A cluster number alone does not establish a correct department or new intent category. Read representative records and assess overlap and usefulness before assigning business meaning.
Distinguish classification from exact matching
Intent classification is not necessarily a check for the category name as a literal string. Evaluate paraphrases and messages with multiple concerns, and do not interpret classification or clustering as automatic reply sending.