Natural Language queries to be answered, with datasource as Elasticsearch

[Vigneshg98]

I went through Haystack and it looks very interesting. I'm trying to build an application that answers the user question based on the dataset they have. Here, I have the source as Elasticsearch.

Sample data in the elastic index.
{ "Order ID" : "ORD-1234", "Order Date" : "5/30/2018", "Ship Mode" : "Standard Mode", "Customer ID" : "CUS-1234", "Customer Name" : "John Doe", "Segment" : "Corporate", "Country/Region" : "United States", "City" : "Parker", "State" : "Colorado", "Region" : "West", "Product ID" : "PRD-1234", "Category" : "Supplies", "Product Name" : "Binders", "Sales" : 22.62, "Quantity" : 2, "Price" : 11.31, "Discount" : 0.7, "Profit" : -15.08 }

Query: Show me the sales that happened in the West region in Standard Mode.

It would be helpful if you could point to the right Reader/Retriever or any samples to achieve my usecase.

[vblagoje]

Seems like you need to explore text-to-SQL models and potentially fine-tune them on this data. Start from https://huggingface.co/microsoft/tapex-large-sql-execution and continue onward.