Integrate sematic search for Azure Open AI on your data

Question

Integrate sematic search for Azure Open AI on your data

lakshmi 821

We are integrating azure open ai with bot framework for gpt feature.

We wanted to use the Azure OpenAI on your data feature and how we can use the sematic search for the private data in azure open ai.

This needs to be configured in azure search ? What is the possible ways to implement sematic search as same as keyword search in azure open ai

Ramr-msft 17,836 Reputation points

2023-11-27T07:25:49.4566667+00:00

Thanks for the question, Azure OpenAI on your Data supports Cognitive Search, you can use our built-in features as well. Did you enable Semantic Search on your search service?

Here is the document for semantic search that could help. https://learn.microsoft.com/en-us/azure/search/semantic-search-overview
lakshmi 821 Reputation points

2023-11-27T08:58:17.2066667+00:00

@Ramr-msft

Thanks for the reply.

One more question , if private data is enabled in the azure open ai, will it fetch the answer which is not in the data. If user is asking a genral question, will open ai fetch the answer from the already trained data set of gpt.

1 answer

Your answer

Ramr-msft 17,836 Reputation points

2023-11-27T07:25:49.4566667+00:00

Thanks for the question, Azure OpenAI on your Data supports Cognitive Search, you can use our built-in features as well. Did you enable Semantic Search on your search service?

Here is the document for semantic search that could help. https://learn.microsoft.com/en-us/azure/search/semantic-search-overview
lakshmi 821 Reputation points

2023-11-27T08:58:17.2066667+00:00

@Ramr-msft

Thanks for the reply.

One more question , if private data is enabled in the azure open ai, will it fetch the answer which is not in the data. If user is asking a genral question, will open ai fetch the answer from the already trained data set of gpt.

Answer 1

lakshmi Thnaks for the details, if you are asking to restrict the answers to just the knowledge you provide, you can do this via the RAG pattern (as part of prompt engineering). it is possible to limit answers to only data sources; however, it's not completely foolproof so responsible implementation, including great human and AI interaction design guidelines are highly recommended. Later models (re: gpt-4) are more steerable and adhere to rules and instructions in the system message better than others.

Share via

Integrate sematic search for Azure Open AI on your data

1 answer

Your answer