From Text to Context: How We Introduced a Modern Hybrid Search
Offered By: EuroPython Conference via YouTube
Course Description
Overview
Explore the journey of implementing a modern hybrid search system in this 46-minute conference talk from EuroPython 2024. Discover how a global online marketplace selling 20 million products annually transitioned from a traditional word-match based approach to a cutting-edge hybrid solution combining BM25 with a semantic vector model. Learn about the challenges faced in both machine learning and engineering aspects, including data pipeline encoding, live service query encoding, and integration with search engines. Gain insights into the system's architecture, implementation strategies for managing latency, and the process of fine-tuning an open-source language model for domain-specific applications. Understand the shortcomings of traditional search methods, the benefits of hybrid search, and the potential for improving user experience and product discoverability in e-commerce platforms.
Syllabus
From Text to Context: How We Introduced a Modern Hybrid Search — Ansgar Gruene, Dharin Shah
Taught by
EuroPython Conference
Related Courses
SPLADE - The First Search Model to Beat BM25James Briggs via YouTube Supercharge eCommerce Search - OpenAI's CLIP, BM25, and Python
James Briggs via YouTube Vector-Based Methods for Similarity Search - TF-IDF, BM25, SBERT
James Briggs via YouTube RAG with Verified Citations - Advanced Inference Techniques
Trelis Research via YouTube Mastering Retrieval for LLMs - BM25, Fine-tuned Embeddings, and Re-Rankers
Trelis Research via YouTube