Can i use NLP (Question answer) on structured data?

9d224d4833094bd482cf · February 6, 2024, 6:48am

We are using Elasticsearch database. We are planning to provide global search with lot of filters. Our data is mostly structured and there are built in relationships. Having many filters on global search can create usability problems.

Is it possible to perform NLP tasks like question/answer or chat bot on structured data? I believe this will help us to get rid of filters and user will be able to search through questions.

Tim_Grein · February 6, 2024, 8:59am

Hi Ashit,

Thank you for your question! There is ongoing development in the (tech-preview) inference API, which will probably have more task types added in the future. There's also built-in NLP support, which also supports question answering models. Does this help?

9d224d4833094bd482cf · February 7, 2024, 5:02am

Hi Tim,
Thanks for prompt response.

I was wondering if it is possible to do following:
User will ask questions like in UI -
How many students are suffering from covid 19?
My index doesn't store values in vector form but I would still like to convert above question into proper elastic query and return result to user.

joemcelroy · February 7, 2024, 10:19am

we have some examples of how you would do this within LangChain.

In this example notebook, you provide a schema for your index and it will transform the question into an BM25 only query.

github.com

elastic/elasticsearch-labs/blob/main/notebooks/langchain/self-query-retriever-examples/chatbot-with-bm25-only-example.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# BM25 and Self-querying retriever with elasticsearch and LangChain\n",
    "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/elastic/elasticsearch-labs/blob/main/notebooks/langchain/notebooks/langchain/self-query-retriever-examples/chatbot-with-bm25-only-example.ipynb)\n",
    "\n",
    "This workbook demonstrates example of Elasticsearch's [Self-query retriever](https://api.python.langchain.com/en/latest/retrievers/langchain.retrievers.self_query.base.SelfQueryRetriever.html) to convert unstructured query into a structured query and we use this for a BM25 example. \n",
    "\n",
    "In this example:\n",
    "- we are going to ingest a sample dataset of movies outside of LangChain\n",
    "- Customise the retrieval strategy in ElasticsearchStore to use just BM25\n",
    "- use the self-query retrieval to transform question into a structured query\n",
    "- Use the documents and RAG strategy to answer the question "
   ]
  },
  {
   "cell_type": "markdown",

This file has been truncated. show original

Behind the scenes, LangChain is providing a prompt to transform the query into a structured query, based on the schema and then translating that structured query into a BM25 query. You could do this process manually, without LangChain.

9d224d4833094bd482cf · February 8, 2024, 11:31am

Hi Joseph,
This is really useful. I will try this out.

Could you please elaborate more on your comment - we can translate into structured query manually without LangChain

General Question - Any recommendation on security measures?

joemcelroy · February 9, 2024, 7:27pm

This article may help give you an understanding of whats happening in self querying Self-querying | 🦜️🔗 Langchain

You could do this yourself without needing LangChain, just means more code to write :). The prompt is:

providing an instruction to transform the given query into a structured query and giving examples of this.
The prompt also provides a schema of the fields available, so the LLM knows what attributes it could use for filtering / searching on

This article may help give context also How to use Elasticsearch to prompt ChatGPT with natural language | Elastic Blog

On security, depends on what security requirements you have. You would need to sanitise the inputs just like you do in any production application.

Joe

9d224d4833094bd482cf · February 16, 2024, 1:09pm

Thank you !
Appreciate your help

system · March 15, 2024, 1:09pm

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.