CODEWITHSUNDEEP
azureazure-cognitive-servicesazure-cognitive-search
userbin
userbin

Reputation: 59

How to use Azure Cognitive Search to find answers to questions from unstructured documents?

I am new to Azure Cognitive Search. I uploaded different types of files in the Azure Blob Storage (.html, .pdf, .xlsx,.docx ..). I want to use the search capabilities of Azure Cognitive search to get a document and an answer/highlights which represents what I am searching about.

For example, I have different files with different fruits as data, and if I ask "where are oranges grown?" then I would expect desired highlighted text with that answer and the document in question. When I search this, I normally get all the documents which has "where", "are" are frequently mentioned but not the result itself. Is there anything I can do to get desired answer?

Another way is to integrate QnA maker with Azure Cognitive Search, which does give question and answers. However, I am searching for an option where I don't use QnA maker. Is it possible to do so? Semantic search does give some result, but it is bit expensive for me.

Another scenario, there is an image in a pdf to make orange juice. I want to get that image as a part of an answer when I search "how to make orange juice?", is it possible to do so?

Upvotes: 1

Views: 1018

Answers (1)

Gia Mondragon - MSFT
Gia Mondragon - MSFT

Reputation: 466

The Q&A built-in experience you're looking for can be achieved by integrating either through Azure Cognitive Service for Language or by the Semantic Search. There is no built-in mechanism that is not through Semantic Search in the service itself to get that specific functionality.

In regard to returning images, you should be able to use skillsets and specifically the Image analysis skill to tag your images and have a URL of the image location as one of the index fields, so you can return in your results. Vector Search is not supported at this time.

I hope this helps.

Upvotes: 0

Related Questions