Skip to content

Synonyms & Taxonomy Tools

Marqo's LLM-based search automatically handles synonyms, brand aliases, misspellings, and regional variations, ensuring users find products regardless of how they search—without requiring manual synonym management.

Overview

Traditional search systems require extensive manual work to maintain synonym lists, handle misspellings, and manage regional language variations. Marqo's LLM Based Ranking and Relevance system automatically understands query intent, recognizes synonyms, handles misspellings, and adapts to regional language differences—all without manual configuration.

How LLM-Based Synonyms Work Better

Automatic Synonym Recognition

The LLM-based system understands that "running shoes", "sneakers", "athletic shoes", and "trainers" refer to the same product category. It recognizes these relationships automatically based on:

  • Semantic understanding: The LLM understands meaning, not just exact word matches
  • Context awareness: It knows when "trainers" means shoes vs. fitness instructors
  • Continuous learning: It learns from customer behavior which terms are equivalent

Misspelling Handling

The system automatically handles common misspellings: - "nkie" → "Nike" - "addidas" → "Adidas" - "iphne" → "iPhone"

This happens automatically without maintaining a misspelling dictionary.

Brand Alias Recognition

Brand name variations are understood automatically: - "Nike Inc" → "Nike" - "Apple Computer" → "Apple" - "Samsung Electronics" → "Samsung"

The LLM recognizes these as the same brand without manual aliasing.

Regional Variations

The system adapts to regional language differences: - US: "sneakers" or "tennis shoes" - UK: "trainers" or "plimsolls" - AU: "runners" or "sneakers"

All understood automatically based on user context and location.

Why LLM-Based Synonyms Are Better

  • No maintenance: No need to manually create and update synonym lists
  • Better accuracy: LLMs understand context and meaning, not just word matching
  • Handles ambiguity: Knows when "trainers" means shoes vs. people
  • Adapts to trends: Learns new terms and variations as language evolves
  • Comprehensive coverage: Handles synonyms you might not have thought of

Fine-Grained Control Through Search Merchandising

While synonyms are handled automatically, you still have fine-grained control through search merchandising:

  1. Navigate to Merchandising → Search Merchandising
  2. Enter a query (e.g., "sneakers")
  3. View results and see how the system interpreted the query
  4. Adjust rankings using boost/bury, pinning, or exclusions if needed
  5. Test variations to ensure different query terms return appropriate results

This gives you the best of both worlds: automatic synonym handling with the ability to fine-tune results for specific queries when needed.

Best Practices

  1. Trust the system: The LLM handles most synonym cases automatically—let it work
  2. Monitor performance: Check analytics to see how well different query variations perform
  3. Use merchandising for edge cases: If a specific synonym isn't working as expected, use search merchandising to adjust results
  4. Test variations: Periodically test different query terms to ensure they return relevant results