RAGChat: Customizing our RAG solution

แชร์
ฝัง
  • เผยแพร่เมื่อ 30 ม.ค. 2025

ความคิดเห็น • 4

  • @shanemickman
    @shanemickman วันที่ผ่านมา

    hey can we use this for microsoft sharepoint documents also? and if so can we use the same document access based on the users (document level wise access for the bot answers )?

  • @markshark337
    @markshark337 8 วันที่ผ่านมา

    Hello there, continuing to watch the sessions of RAG Deep Dive at more European times 😊. Did you make any evaluation of chunking, vectorization and indexing when using python script vs Integrated vectorization? I mean are there any drawbacks just from looking at answer quality point of view?

    • @PamelaFox
      @PamelaFox 7 วันที่ผ่านมา +1

      I compared those two options in the follow-up talk about AI Search: aka.ms/ragdeepdive/aisearch/slides
      Check out slide 30 for a comparison. There's definitely a difference in splitting, like the biggest issue of no page numbers for the integrated vectorization splitter, but also they're just different splitting libraries, so you'll see different results.

    • @markshark337
      @markshark337 3 วันที่ผ่านมา

      @@PamelaFox thanks, for this I know. Let me ask in simpler way: is there any difference in the quality of results/answers when using Integrated Vectorization or manual indexing code?