The Capabilities of Semantic Search
summary
Compares symbolic indexing (Elasticsearch/OpenSearch) with vector-based retrieval and shows a working pipeline using the Jina framework to encode items into embeddings (300–500 dims) and match with cosine similarity. Demonstrations include text and image queries, a live demo at jina.ai/now, and a small project that searched 5 government documents (~80 pages each) to extract relevant passages with example similarities around 0.29. Discusses foundation models, self-supervision (no labels required), and the need to monitor and retrain for domain shift.
AI-written summary · about
chapters (9)
- 0:00Intro
- 0:30What is Semantic Search
- 3:03Semantic Search
- 4:41Matching
- 7:22Embedding
- 9:27China
- 10:14Ongina
- 14:53Questions
- 20:11Selfsupervision
related talks
30mThe Unlimited Potential of Neural Search to Unlock the New Way of Data Comprehension
32mBuilding semantic search using Deep Learning
13mA Beginner’s Guide to Semantic Search
48mUnlocking Next-gen Semantic Search With OpenSearch 3.x
40mVector Search Made Simple: Getting Started with OpenSearch for AI Applications
32mRetrieval Augmentation and Semantic Search at Scale
31mAdvanced search for your e-commerce website
01h 30mBuilding a Search Engine