Event banner
Event details
116 Comments
- note New OpenAI LZ architecture published - https://techcommunity.microsoft.com/t5/azure-architecture-blog/azure-openai-landing-zone-reference-architecture/ba-p/3882102
- For some new to AI and Cognitive Search, what skills should you learn to get a job ? are those skills on MS Learn :-)? Thank you in advance for guidance!!!! 🙂
Microsoft does have some relevant modules on MS Learn to help you boost your skills. Here are a few courses that you could consider:
- Introduction to Azure Cognitive Search - Training | Microsoft Learn
- Introduction to Azure OpenAI Service - Training | Microsoft Learn
Jason Jensen here from south-eastern Washington state. I'm interested in information or resources on what the Retrieval Augmented Generation pattern with Cognitive Search and vector search looks like end-to-end. Specifically comparing it to the existing samples available in GitHub would be extra useful.
- liamca-msft
Microsoft
Jason, I think this blog post that Pablo Castro wrote is the best resource to get started here and to learn more about RAG and the Retrieve Augment Generate pattern. From there, the demo code which can be found here. is a great more technical resource.
Thanks liamca-msft , that is the pattern I have been following in my experimentation using the sample code from GitHub as well. How will vector search change or improve upon this?
- When will the documentation on the Semantic Index and how to use it be released?
- mike_carter_msft
Hi Pat. You can find documentation about semantic search in Azure Cognitive Search here: https://learn.microsoft.com/en-us/azure/search/semantic-search-overview When you say "Semantic Index", is this what you are referring to?
- thx. No Semantic index is according to the docs is what Msft Co-Pilots will be using. My understanding is it will be avaiable for customer use (E3 and E5) as well, but have not seen any documentation. My working assumption is it will be exposed via a graph connector perhaps.
- is there a live link or video or presentation? or is it focused on the questions above and below for the event?
- Like you I was also thinking this was a streamed event. I can't see any video feed.
- There is no video or audio feed to stream. All text!
- Hello - there is no audio or video component for this event. It is similar to a Reddit AMA in that it is all text-based. Yes, it is focused on the questions above and below for the event. Feel free to post any questions if you have them!
- Hi i'm connecting from California in the bay area. It seems vector search will usually outperform a text-based semantic search in terms of accurate results. Do you have any hard data showing a side by side comparison of these 2 search methods?
- yahnooshHi Glenn. Vector search and Semantic search are somewhat orthogonal but aim to serve the same function – improve the quality of the search results. In this context, Vector search is used to improve recall – the number of relevant results returned by a search query. Semantic search improves ranking by bringing up the most relevant documents to the top of the results list. Vector search and Semantic search can be used independently but work best when used together. Note, Semantic search today works only for search requests that include a text query. The effectiveness of each method will depend on your scenario. For example, Vector search is only as good as the model you use to vectorize the data and you should refer to the model benchmarks to understand its strengths and weaknesses. Semantic search leverages Bing models trained on the web corpus, so it’s very effective for data that’s not domain-specific, but it’s best to test yourself on your data (Semantic search is free for the first 1000 requests). Please let me know if this answers your question.
- very helpful, thank you!
- thegovind
Will "auto-vectorization" with an embedding model of choice (AOAI's ada or HF's sentence transformers), including intelligent splitting/chunking, be an option for any of Azure's persistence providers or ACS?
- gia_mondragonPlease stay tuned for such functionality in the upcoming months. It is currently in our roadmap.
robertlee-msftCould you please share what elements you would consider to be part of "intelligent splitting/chunking"? Would this be based on token count, sentence boundaries, paragraph boundaries, or other considerations?
- I'd love to gain some insight into the pricing model of vector search. Based on the limited information available on the semantic feature in Cognitive Search, I'm guessing there will be some base cost of ~$500 per month. Why the base cost component? Also, what will be the price be approximately per token/document ingested? Thanks!
- robertlee-msft
Adding to Liam's answer, using semantic search along with vector search is not a requirement to use vector search features itself. Customers may often find that semantic search will improve the relevance and ranking of the search results, so you may wish to perform testing on whether to use semantic search if you are more sensitive to the base cost.
If you are only using vector search, there are no additional costs to use this feature within your pre-existing search service. Please note the service limits and restrictions outlined in these pages below:
Vector limits: https://learn.microsoft.com/en-us/azure/search/search-limits-quotas-capacity#vector-index-size-limits
Vector availability and pricing: https://learn.microsoft.com/en-us/azure/search/search-limits-quotas-capacity#vector-index-size-limits
- liamca-msftThat is correct Brennan, unfortunately (or fortunately), the semantic model leverages GPU's that has a cost associated with it. We do understand that it makes more sense for some customers to have a more gradual ("pay per usage") type model as opposed to having the $500 cost for a large number of transactions. Please keep an eye out for that. It is important to note that for the semantic search capability, there is no cost for ingesting data as it is purely a reranking model. More details can be found here: https://learn.microsoft.com/en-us/azure/search/semantic-search-overview
Welcome to the Azure Cognitive Search Ask Microsoft Anything!
This live hour gives you the opportunity to ask questions and provide feedback directly to the Azure Cognitive Search team.
Please post any questions in a separate, new comment thread.
To start introduce yourself below and tell us where you're logging in from!
- liamca-msft
Hi everyone, my name is Liam Cavanagh and I am the Group Product Manager for Azure Cognitive Search and sit within the engineering group. I really look forward to your questions as well as any feedback you might have. Hopefully you have also seen our public preview announcement of Vector (and Hybrid) Search.
- yahnooshHi! My name is Janusz Lembicz, I’m an engineer currently leading the Vector Search work in Azure Cognitive Search. Thanks for your interest in our service!
- mike_carter_msftHi! I'm a program manager on the Azure Cognitive Search team. I'm looking forward to answering your questions today.
- Hello. The Add to Calendar .ics file seems to suggest that this will be happening 26 Jul from 2 to 3pm. My understanding is that it's 25 Jul 9 to 10am.
- It's working correctly for me! 9am-10am July 25 is Pacific Time Zone. Are you in a different time zone?
- I'm on the Melbourne/Sydney time zone.
