Using LLMs to improve your Search Engine
summary
Covers concrete claims and demos: using LLMs to extract query intent, judge query–item relevance, generate user queries from product titles, and fine‑tune models on an e‑commerce corpus. Shows code and notebooks (GitHub + Google Colab), uses LangChain + Elasticsearch for retrieval and orchestration, evaluates on an Amazon query‑item dataset (>1M annotated pairs), compares GPT‑4, PaLM and Llama 7/13/70B, and demonstrates a LoRA fine‑tuned Llama 7B that produces usable attribute extraction and relevance labels. Cost, latency and caching tradeoffs are discussed.
AI-written summary · about
chapters (6)
- 0:00Improving search engines with LLMs
- 26:37E-commerce search challenges
- 39:59Large language models
- 1:02:39Prompting for query understanding
- 1:20:39Fine-tuning LLMs for e-commerce
- 1:28:23Practical applications of LLMs in search
related talks
46mLLM—Search Engine Killer?
49mLLM—Search Engine Killer?
21mTransforming search and discovery using LLMs
24mConversational Search with LLMs
51mQuery Understanding With Large Language Models: Techniques and...
29mUsing LangChain and LlamaIndex to customize LLMs...
37mExpanding the Capabilities of LLMs Through Community Collaboration
33mOpening Keynote: LLMs, between fantasy and reality - the poin...