For a table projection in a knowledge store (skillset definition), the table node must define:
tableName : the table to project into (physical expression).
generatedKeyName : the name of the system-generated key column.
source : an object-shaping expression (projection path) describing what document content flows into the table.
dataSource/dataSourceConnection are service-level objects used by the indexer, not properties of a table projection node.
References
Knowledge store and projections schema: table projections require tableName, generatedKeyName, and source in the skillset JSON.

Explanation:

Box 1: No
Example: Greet a user whose name is stored in `user.name`
- ${ welcomeUser(user.name) }
Example: Greet a user whose name you don ' t know:
- ${ welcomeUser() }
Box 2: No
Greet(User) is a Send a response action.
Box 3: Yes
Reference:
https://docs.microsoft.com/en-us/composer/how-to-ask-for-user-input

You need to identify industry-specific technical terms in documents. That is a classic use case for Custom Named Entity Recognition (custom NER) , which lets you build a model to extract domain-specific entities (for example, chemical names, part numbers, contract IDs) from unstructured text. Among the options, only custom NER is designed to recognize entities unique to your domain.
Why not Key Phrase Extraction? It returns the main concepts/talking points of text, not typed, domain- specific entities.
Why not CLU? Conversational Language Understanding is for intent and entity extraction from user utterances in conversations , not for mining technical terms from documents. Microsoft Learn Why not Language Detection? It detects the language of the text only.
Custom NER fits: Purpose-built to extract domain-specific entities with minimal code; you label a small dataset and train in Language Studio.
Microsoft References:
Custom Named Entity Recognition overview - purpose is extracting domain-specific entities . Microsoft Learn Transparency note for custom NER - emphasizes domain-specific extraction. Microsoft Learn Key Phrase Extraction overview - identifies main concepts , not domain entities. Microsoft Learn CLU overview - intent/entity for conversational scenarios. Microsoft Learn

To generate complete sentences describing an image (for accessibility for visually impaired users), you must call the Describe Image API.
describeImageInStreamAsync analyzes the content of the image and generates human-readable descriptions in natural language sentences.
readInStreamAsync is for OCR (extracting text).
tagImageInStreamAsync is for generating tags (keywords), not full sentences.
analyzeImageByDomainInStreamAsync is for domain-specific models (e.g., celebrities, landmarks).
Reference: Computer Vision Describe Images

Active learning is a technique of machine learning in which the machine learned model is used to identify informative new examples to label. In LUIS, active learning refers to adding utterances from the endpoint traffic whose current predictions are unclear to improve your model.
Reference:
https://docs.microsoft.com/en-us/azure/cognitive-services/luis/luis-glossary