Learn Generative AI, Prompt Engineering, and LLMs for Free
The Most Addictive Python and SQL Courses
Overview
AI, Data Science & Cloud Certificates from Google, IBM & Meta — 40% Off
One plan covers every Professional Certificate on Coursera. 40% off Coursera Plus Annual.
Unlock All Certificates
This 30-minute tutorial from Trelis Research explores the differences between keyword and vector search methodologies for information retrieval. Learn the pros and cons of each approach, with detailed explanations of how vector search embeds semantic meaning and how keyword search focuses on exact matches. Discover techniques for handling challenges like long documents and new/infrequent words, and explore how LLMs can optimize queries. The video covers practical implementations including Postgres libraries for search and how to implement BM25 in Postgres using tensorchord's vectorchord-bm25. Gain insights into hybrid search techniques that combine the strengths of both methods for more effective information retrieval systems. The presentation includes timestamps for easy navigation through topics ranging from basic search concepts to advanced implementation strategies.
Syllabus
00:00 Introduction to Vector Search
00:34 Pros and Cons of Keyword vs. Vector Search
02:13 Understanding Vector Search
03:35 Keyword Search Explained
06:17 Comparing Search Techniques
07:40 Handling Long Documents
09:31 New and Infrequent Words
10:50 LLMs and Query Optimization
12:17 The Importance of Pagination
13:22 Postgres Libraries for Search
20:18 Implementing BM25 in Postgres with tensorchord's vectorchord-bm25
23:50 Hybrid Search Techniques
29:58 Conclusion and Further Resources
Taught by
Trelis Research