6 Using context to learn domain-specific language

 

This chapter covers

  • Classifying query intent
  • Query-sense disambiguation
  • Identifying key terminology from user signals
  • Learning related phrases from user signals
  • Learning misspellings and alternate term variations from user signals

In chapter 5, we demonstrated both how to generate and use a semantic knowledge graph (SKG) and how to extract entities, facts, and relationships explicitly into a knowledge graph. Both techniques rely on navigating either the linguistic connections between terms in a single document or the statistical co-occurrences of terms across multiple documents and contexts. You learned to use knowledge graphs to find related terms, and how those related terms can integrate into various query-rewriting strategies to increase recall or precision.

In this chapter, we’ll dive deeper into understanding query intent and the nuances of using different contexts to interpret domain-specific terminology in queries. We’ll start by exploring query classification and then show how those classifications can be used to disambiguate queries with multiple potential meanings. Both approaches will extend our use of SKGs from the last chapter.

6.1 Classifying query intent

6.2 Query-sense disambiguation

6.3 Learning related phrases from query signals

6.3.1 Mining query logs for related queries

6.3.2 Finding related queries through product interactions

6.4 Phrase detection from user signals

6.4.1 Treating queries as entities

6.4.2 Extracting entities from more complex queries

6.5 Misspellings and alternative representations

6.5.1 Learning spelling corrections from documents

6.5.2 Learning spelling corrections from user signals

6.6 Pulling it all together

Summary