Parent-child document retrieval in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Parent-Child Document Relationships

In a parent-child document retrieval system, what is the main purpose of establishing a parent-child relationship between documents?

A To randomly assign documents to unrelated categories

B To merge all documents into a single flat list for faster search

C To prevent any document from being indexed in the system

D To enable efficient retrieval of child documents based on parent document queries

💡 Hint

Think about how grouping related documents helps in searching.

Parent-child relationships allow the system to retrieve child documents when querying the parent, improving search relevance and structure.

❓ Predict Output

intermediate

Output of Parent-Child Query in Elasticsearch

Given the following Elasticsearch query to find child documents with a parent of type 'article' having title 'AI Basics', what is the expected output count?

Prompt Engineering / GenAI

POST /my_index/_search
{
  "query": {
    "has_parent": {
      "parent_type": "article",
      "query": {
        "match": { "title": "AI Basics" }
      }
    }
  }
}

A Number of child documents whose parent article's title matches 'AI Basics'

B Number of parent documents with title 'AI Basics' only

C All documents in the index regardless of parent-child relation

D Zero, because 'has_parent' queries are invalid

💡 Hint

Focus on what 'has_parent' query does in Elasticsearch.

'has_parent' returns child documents whose parent matches the query condition, so output count is child docs linked to parents titled 'AI Basics'.

❓ Model Choice

advanced

Choosing the Best Model for Parent-Child Document Retrieval

Which model architecture is best suited for learning hierarchical relationships in parent-child document retrieval tasks?

A Recurrent Neural Networks (RNNs) without attention mechanisms

B Graph Neural Networks (GNNs) that model document nodes and edges

C Convolutional Neural Networks (CNNs) designed for image data

D Simple feedforward neural networks ignoring document structure

💡 Hint

Consider models that can represent relationships between entities.

GNNs naturally model nodes and edges, making them ideal for hierarchical parent-child document relationships.

❓ Metrics

advanced

Evaluating Retrieval Performance for Parent-Child Queries

Which metric best measures the quality of retrieving child documents given a parent document query in a parent-child retrieval system?

A Mean Average Precision (MAP) to evaluate ranking quality of retrieved child documents

B Root Mean Squared Error (RMSE) used for regression tasks

C BLEU score used for machine translation quality

D Confusion matrix used for classification accuracy

💡 Hint

Think about metrics that evaluate ranked retrieval results.

MAP measures how well the system ranks relevant child documents higher, which is key in retrieval tasks.

🔧 Debug

expert

Debugging a Parent-Child Retrieval Query Returning No Results

You run a parent-child query in your system but get zero results, even though you know matching documents exist. Which is the most likely cause?

A The system automatically merges parent and child documents, so separate queries are invalid

B The query syntax is correct but the documents have no fields

C The parent and child documents are indexed in different indices without a proper join setup

D The child documents are always returned regardless of parent query

💡 Hint

Consider how parent-child relationships depend on indexing structure.

Parent-child queries require both document types in the same index with join relationships; separate indices cause no matches.