Knowledge Graph-Guided Search

User Intent

"How can the knowledge graph improve my search results? Show me entity expansion and graph-aware retrieval."

Operation

SDK Methods: queryObservables() + queryContents() combined Concept: Use entity relationships to enhance search Use Case: Graph-enhanced information retrieval

Prerequisites

• Knowledge graph with entities

• Search queries

• Understanding of entity relationships

---

Key Concepts

1. Entity Expansion

Expand search terms using entity variants:

// User searches "Kirk"
// System expands to:
// - "Kirk Marple"
// - "K. Marple"
// - [email protected]

2. Relationship-Aware Search

Use entity relationships to broaden results:

// Search "Graphlit team"
// 1. Find Graphlit (Organization)
// 2. Find all People at Graphlit
// 3. Return content mentioning any team member

3. Entity Disambiguation

Use context to identify correct entity:

// "Apple" could be company or fruit
// Knowledge graph helps disambiguate

---

Complete Code Example (TypeScript)

import { Graphlit } from 'graphlit-client';
import { ObservableTypes } from 'graphlit-client/dist/generated/graphql-types';

const graphlit = new Graphlit();

async function graphGuidedSearch(query: string) {
  console.log(`Searching: "${query}"\n`);
  
  // Step 1: Find entities matching query
  const entities = await graphlit.queryObservables({
    search: query
  });
  
  console.log(`Found ${entities.observables.results.length} matching entities\n`);
  
  // Step 2: For each entity, find related entities
  const allEntityIds = new Set<string>();
  entities.observables.results.forEach(e => allEntityIds.add(e.observable.id));
  
  for (const entity of entities.observables.results.slice(0, 3)) {
    const related = await graphlit.queryContents({
      filter: {
        observations: [{ observable: { id: entity.observable.id } }]
      }
    });
    
    // Extract related entities
    related.contents.results.forEach(content => {
      content.observations?.forEach(obs => {
        allEntityIds.add(obs.observable.id);
      });
    });
  }
  
  console.log(`Expanded to ${allEntityIds.size} entities via relationships\n`);
  
  // Step 3: Search content mentioning any of these entities
  const expandedResults = await graphlit.queryContents({
    search: query  // Semantic search
    // Plus entity filter (if needed)
  });
  
  console.log(`Results: ${expandedResults.contents.results.length}`);
  
  return expandedResults.contents.results;
}

await graphGuidedSearch("Graphlit");

---

Benefits

Better Recall: Find variations and related mentions Entity Resolution: "Kirk" expands to "Kirk Marple" Context Awareness: Relationships provide context Disambiguation: Choose correct entity meaning Richer Results: Include related entities

---

Patterns

Pattern 1: Team Search

"Show me Company X team" → Find all people at Company X

Pattern 2: Topic Expansion

"AI research" → Find research + related papers + authors

Pattern 3: Temporal Context

Entity mentions over time with relationship context

---

Developer Hints

• Entity resolution improves recall

• Graph relationships add context

• Combine with semantic search

• Better than pure keyword search

• Cache entity expansions

---