I am Sumit and I'm trying to prototype a feature for my product and need ML/NLP guidance. I don't need a bulletproof solution, just something to work. My product, dwata.com, aims to make database queries available to non-technical people without SQL. I've seen recent YC companies launch to solve similar issues (BaseDash being most recent). I'm trying to tackle consumption and exploration of data more than modification.

### What I have A Python/SQLAlchemy based query builder which infers relations, column types, cardinality and does the best suited JOINs (1-1, M-1) or embedded queries (1-M or M-M). The intent of the user matters (expressed visually), and if separate SQL is better, the software should do that.

### The ML/NLP part What I want is to be able to answer a question like "show me latest orders of red bags". Assuming that "Red" is a Color value in a Product Attribute table and Bag is a Product Category. This is all made up but I think it is easy to understand what I am trying minimally.

### What I think is a start Generate English queries for a DB. Like "show me content with tag travel", "show me latest orders". I can use some textual templates to generate permutations of the data and relations. Then get an ML based system to learn these and corresponding correct SQL. Now, as I understand, if I throw a new English query at this learning system, then I should get a new SQL.

### What I need If you are experienced with ML/NLP and are willing to give me a short set of bulletpoint that you feel would get me to a working prototype, it would really help me. I have a very basic idea of NLP, ML but have not worked with them. I need a kick-start.

Note: You can see the current, very basic, query builder here - https://github.com/brainless/dwata/blob/develop/backend/utils/query_builder.py

I would be tremendously thankful if you can spare a little time to guide me, thanks. Sumit Datta