Enabling the world to Search by Meaning
summary
ZIR AI presents a semantic search engine that maps queries and documents into a unified meaning space using word vectors, transformers and unsupervised pretraining (BERT-style) and runs as Search-as-a-Service with Aerospike for storage, indexing and low-latency serving. Demo shows cross-language matches with no keyword overlap. Reported gains include 27% relative jump in R@1 from transformers and operational metrics like p99 <10ms using Aerospike; requests include reservoir sampling and vector proximity lookups.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:24Welcome
- 1:50What is this number
- 3:20Problems with keyword search
- 6:30Knowledge Graphs
- 7:15Neural Rank Demo
- 10:10Why
- 11:30One Hot Vector
- 12:38Transformers
- 14:42Unsupervised Pretraining
- 16:25Applications
- 16:45Aerospike
- 17:34How we use Aerospike
- 18:04Performance
- 19:17Feature Requests
- 20:25Summary
related talks
25mSemantic Search
31mAdvanced search for your e-commerce website
13mThe State of AI Powered Search and Retrieval
30mBeyond Keywords: Building an AI-Powered Search That Reads Your Mind
30mSearch on Steroids — A Story of Semantics
39mAI-powered Semantic Search; A story of broken promises?
30mSemantic Search and Neural Information Retrieval
47mBoosting AI With OpenSearch: Vector Stores, Similarity Search, and Retrieval-Augme...