I have a dataset in csv format. Most columns have numbers in them. The data is mostly number based, with little text.

A sample of my dataset:

enter image description here

I want to build a semantic search engine that can answer questions like the following:

  1. Which user has executed the maximum number of tests?
  2. On what date did a particular user execute tests?
  3. Which date observes the maximum number of failed tests?

I tried Amazon Kendra, Azure Cognitive Search, ChatGPT, Bard, built my own semantic search engine from scratch, but none of them are performing well on numeric data. They're all more text based.

One idea I have is, if I change the numbers to text, will the models be able to predict well?

So for my requiement, what would you suggest me to try? Please shed some light on this.


1 Answer 1


You could feed the LLM a description of the file format and then request it to generate a piece of code to extract the information you want, for instance, in Python. Then, you would run the generated code on your data to compute the actual answer.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.