ground-truth-design

A ground-truth query establishes a known-answer test for semantic search, enabling deterministic evaluation of retrieval quality against curriculum content. It uses a known-answer-first approach to validate that search results align with expected educational materials.

Design ground-truth queries by following a structured process from content discovery to query evaluation, using bulk data and predefined relevance criteria. You run searches and validate top results against known answers using a standardized scoring system to measure search quality.

Yes, you can use bulk data to assess search relevance for the Oak curriculum. The workflow supports loading bulk data and applying predefined relevance criteria to evaluate top results, enforcing a structured process to validate search quality at scale.

Yes, the methodology targets professional UK teachers to verify that search results align with curriculum content. It requires teacher input to apply predefined relevance criteria and use the standardized scoring system for evaluating top search results.

The best way to iteratively refine queries is to use the structured workflow that moves from content discovery to query evaluation. By applying predefined relevance criteria and a standardized scoring system to top results, you can continuously adjust queries to improve measurement accuracy.

A known-answer-first approach is needed for semantic search evaluation because it establishes deterministic tests for retrieval quality. By defining ground-truth queries with known correct answers, you can accurately measure whether the search service returns relevant curriculum content.