Azure AI Search
An Azure search service with built-in artificial intelligence capabilities that enrich information to help identify and explore relevant content at scale.
994 questions
How we can use the "@search.score" and "@search.rerankerScore" response in ai search api call
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(131.20000000000002, 26%, 22%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EL%3C/text%3E%3C/svg%3E)
lakshmi
716
Reputation points
Hi Team,
We are using ai search document Nuget package to find the confidence score and the file paths which has generated the answer. Below is a sample question I have tried in AI search index search query,
Tried the same questions in Open ai studio and below is the result,
If you verify both the results, in open ai its showing only the exact file which has the data and search client shows all most all files which has this keyword.
Using score or rankings core we are unable to filter it out whether the query is actually found in that specific document.
Same we are seeing in Open ai package response also where citations contain may of the documents details even if the actual data is not there also,
Is there any possible ways to fetch the exact confidence score and files which contain the query. whether its possible to use score and rank score value. As in this example case the document which has data and not has data also has a score with 2.0..
{count} votes
-
brtrach-MSFT 16,346 Reputation points Microsoft Employee2024-05-30T05:31:44.6066667+00:00 @lakshmi The "@search.score" and "@search.rerankerScore" are the relevance scores that are returned by Azure Cognitive Search. The "@search.score" is the relevance score of the document based on the query, and "@search.rerankerScore" is the relevance score of the document based on the re-ranking algorithm.
These scores are not absolute measures of confidence. They are relative to the specific query and the set of documents being searched. Therefore, a document with a score of 2.0 does not necessarily mean that it’s twice as relevant as a document with a score of 1.0.
To fetch the exact confidence score and files which contain the query, you can use the "highlight" feature of Azure Cognitive Search. The "highlight" feature returns the fragments of the document that match the query. You can use this feature to identify the exact location of the query in the document.
Here is an example of how you can use the "highlight" feature in your search query:
searchThis will return the documents that match your query and the fragments of the document that match the query in the "your_field" field.
Sign in to comment
Sign in to answer