Azure AI Search with Varying Verbiage

Question

Azure AI Search with Varying Verbiage

Keith Ho 25

Currently, I'm using the Azure AI Search in conjunction with my LLM. I've created an index and uploaded documents into it. I have a Chabot that answers questions based on the information inside the index.

When I ask the Chatbot about "Frozen Macess", the Azure AI Search is able to search through the index for documents that contain those exact words. However, when I change the verbiage to "freezing" instead of "frozen", the AI Search can't find the document it did previously. I'm using the Azure AI Search through a vector store as a retriever with the search type set as "similarity_score_threshold."

Is there a way I can get "freezing Macess" to return the same documents as "Frozen Macess"? Is there a way to retrieve documents when the verbiage is slightly different than the exact wording that is on the documents that I am uploading into the Azure Search Service Index?

Accepted answer

0 additional answers

Your answer

Answer 1

Pavankumar Purilla 4,280 Microsoft External Staff

Hi Keith Ho,

To improve Azure AI Search's ability to handle queries with varying wording, you can enable semantic search, which retrieves documents based on meaning rather than exact keywords. This ensures that queries like "freezing Macess" can match documents containing "Frozen Macess." Additionally, using synonym maps allows you to define relationships between similar terms, ensuring that synonyms or related words are recognized during searches. Applying stemming or lemmatization through language analyzers can further normalize words to their root forms, improving query-document matching.

Lowering the similarity_score_threshold can help retrieve documents that are semantically similar but not exact matches. Preprocessing queries with NLP tools to normalize terms or expand synonyms before sending them to Azure AI Search can also enhance retrieval accuracy.
Finally, combining vector-based search with keyword-based search in a hybrid approach ensures robust results, as vector search focuses on semantic meaning while keyword search ensures exact matches.

Hope this helps. Do let us know if you have any further queries.

If this answers your query, do click Accept Answer and Yes for was this answer helpful.

Keith Ho 25 Reputation points

2025-03-10T22:40:09.93+00:00

I'm trying to set my self.vector_store.as_retriever to use the semantic_hybrid search. I have created a semantic configuration on my Azure AI Search index but I'm running into this error:
"The parameter semanticConfiguration' must not be empty. Alternatively, set a default semantic configuration for this index."

self.vector_store = AzureSearch(....)

retriever = self.vector_store.as_retriever(search_type="semantic_hybrid", k=k, semanticConfiguration="my-semantic-config")

Edit: I was able to resolve this error.

self.vector_store = AzureSearch(azure_search_endpoint=<endpoint>, azure_search_key=<search-key>, index_name=self.index_name, embedding_function=self.embeddings_function, semantic_configuration_name="my-semantic-config"

retriever = self.vector_store.as_retriever(search_type="semantic_hybrid", k=k, semantic_configuration_name="my-semantic-config")
Pavankumar Purilla 4,280 Reputation points Microsoft External Staff

2025-03-11T17:28:52.36+00:00

Hi Keith Ho,
Following up to see if the above answer was helpful. If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.

Share via

Azure AI Search with Varying Verbiage

0 additional answers

Your answer