"Search" sounds like an auxiliary function - until your users can’t find what they’re looking for. Whether they’re trying to read, buy, or make a decision, good search is critical. AI tools like text embeddings and large language models promise smarter search systems that go beyond keywords to capture user intent and preferences. Yet, making that leap isn’t just a matter of swapping your index for a vector database. In this talk, we’ll share lessons learned from building **PaperFinder**, an AI-powered research assistant developed at AI2. We’ll cover the extra steps that can make a difference, from rephrasing to reranking, and more. Finally, we discuss how the same patterns can be applied in other fields, such as e-commerce or recommendation systems. [1] PaperFinder: https://paperfinder.allen.ai/ [2] Allen Institute for AI: https://allenai.org/
Room: Room 3
Tue, Oct 28th, 11:50 - 12:20