My Weird Prompts

Knowledge Graphs vs SQL: How Custom Relationships Change Retrieval

28 min · 18 de jun de 2026
Portada del episodio Knowledge Graphs vs SQL: How Custom Relationships Change Retrieval

Descripción

Most databases only tell you that two things are connected. Graph databases tell you how — and that distinction changes everything about retrieval. In this episode, we explore how custom relationship types like EXPLORES, REFUTES, and BUILDS_ON turn a podcast archive from a flat list into a navigable knowledge structure. We walk through concrete Neo4j schema design, show why embeddings alone collapse important nuance, and explain the vector-first, graph-second pattern that combines semantic search with structural precision. If you've wondered how to move beyond "find me similar documents" to "find me what's structurally relevant," this episode walks through the exact modeling decisions that make it work — including the pitfalls of over-taxonomizing your edge types.

Comentarios

0

Sé la primera persona en comentar

¡Regístrate ahora y únete a la comunidad de My Weird Prompts!

Empezar

2 meses por 1 €

Después 4,99 € / mes · Cancela cuando quieras.

  • Podcasts exclusivos
  • 20 horas de audiolibros / mes
  • Podcast gratuitos

Todos los episodios

200 episodios

Portada del episodio Knowledge Graphs vs SQL: How Custom Relationships Change Retrieval

Knowledge Graphs vs SQL: How Custom Relationships Change Retrieval

Most databases only tell you that two things are connected. Graph databases tell you how — and that distinction changes everything about retrieval. In this episode, we explore how custom relationship types like EXPLORES, REFUTES, and BUILDS_ON turn a podcast archive from a flat list into a navigable knowledge structure. We walk through concrete Neo4j schema design, show why embeddings alone collapse important nuance, and explain the vector-first, graph-second pattern that combines semantic search with structural precision. If you've wondered how to move beyond "find me similar documents" to "find me what's structurally relevant," this episode walks through the exact modeling decisions that make it work — including the pitfalls of over-taxonomizing your edge types.

18 de jun de 202628 min