Modern Approaches to Product Matching Using LLMs
summary
The talk explains how large language models (LLMs) are combined with traditional machine-learning algorithms to solve product matching (determining whether two product listings are the same) for marketplaces like Wildberries. It gives concrete examples of using LLMs to extract product attributes and then use those attributes in downstream matching workflows.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 2:55Classical matching problems
- 4:02Text matching with LLMs
- 10:53Image matching with VLMs
- 18:35Model optimization for production
- 24:14Implementation details
- 31:25LLM approach pros and cons
- 32:48Conclusion and Q&A
related talks
20mBeyond Relevance: Using LLMs to Transform E-commerce Search into Product Discovery Engines
22mMachine Learning: The Engine Behind Modern E-commerce Search
01h 30mUsing LLMs to improve your Search Engine
25mTransforming E-Commerce with LLMs: 3 Real-World Use Cases
28mRetrieval-Augmented AI Agents for Entity Matching at Scale
21mTransforming search and discovery using LLMs
24mConversational Search with LLMs
49mLLM—Search Engine Killer?