liveBook · Manning

Related query discovery

As explained in AI-Powered Search

Related query discovery is the process of identifying query terms that are associated through shared user behavior, clicked products, or other retrieval signals.

Related query discovery identifies terms connected by shared product interactions. Queries that lead users to click the same product become evidence for relatedness. section 6.3.2

A product-interaction pipeline proceeds by:

  • Using clicked-query popularity as the master query list before applying PMI2 and composite scoring. section 6.3.2

Shared-product relationships

The resulting pairs include “laptops”–“laptop,” “tablets”–“tablet,” “tablet”–“ipad,” and “ipad”–“apple.” Product aggregation produces 1,579,710 candidate pairs and preserves fine-grained variants such as “ipad 2,” “ipad2,” and “i pad.” section 6.3.2

How AI-Powered Search covers it

How deep it goes
Explains signal joins, pair aggregation, popularity counts, PMI2, and composite scoring, while omitting the scoring calculations.
Examples
Uses shared-product pairs such as “tablet”–“ipad,” “ipad”–“apple,” and spelling variants of “ipad.”
What it assumes
Assumes query and click signals with query IDs, users, targets, and signal types; teaches their aggregation into relatedness evidence.
What you can do after
Build it from raw signals with Spark SQL, then produce scored related-query pairs for downstream phrase processing.
How it connects
Connects query logs, product interactions, PMI2, composite scoring, and misspelling detection.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of AI-Powered Search. The section links point to the parts of the book it draws on.