Uploaded October 2026 | Updated November 2026, 2 days ago
I wanted to re-listen to books from my audiobook library, but its search only matches words in the title. After 15 years of listening, I rarely remember exact titles. So I built a search app on Db2 12.1.5 using its text search and vector search features.
In this video I explain four ways to search, then compare them live on one query: "how to build better habits."
• Keyword search: finds exact words, and where synonym expansion helps
• Semantic search: how vector embeddings match meaning, not words
• Hybrid search: why each retriever has a blind spot, and how combining them covers both
• Reranker: a specialized model that re-scores the top results from stage 1 retrieval, and why you don't run it on everything
Code and setup steps (Db2 AI Cookbook): github.com/IBM/db2-ai-cookbook/tree/main/03-hybrid-search
I'm Shaikh Quader, AI Architect at IBM Db2.
#Db2 #HybridSearch #VectorSearch
I wanted to re-listen to books from my audiobook library, but its search only matches words in the title. After 15 years of listening, I rarely remember exact titles. So I built a search app on Db2 12.1.5 using its text search and vector search features.
In this video I explain four ways to search, then compare them live on one query: "how to build better habits."
• Keyword search: finds exact words, and where synonym expansion helps
• Semantic search: how vector embeddings match meaning, not words
• Hybrid search: why each retriever has a blind spot, and how combining them covers both
• Reranker: a specialized model that re-scores the top results from stage 1 retrieval, and why you don't run it on everything
Code and setup steps (Db2 AI Cookbook): github.com/IBM/db2-ai-cookbook/tree/main/03-hybrid-search
I'm Shaikh Quader, AI Architect at IBM Db2.
#Db2 #HybridSearch #VectorSearch










