Semantic search on help documents

Elastic Stack Elasticsearch
1

Hello,
We have a help documents on our website for our web application. This help documents on our websites are well documented (with text and images). It has a list of Topics and each topic has its own url with sub topic in it. We would like to build a UI where user can search for question (mostly how to questions) and we would like to present content from help center.

I am thinking about solution with following:

  1. Vectorize each topic and sub topic into elastic document
  2. Vectorize user questions and store them into Elasticsearch database.
  3. Retrieve similar questions and find relevant data from elastic
  4. Use LLM to summarize (including text and images).

Questions:
Is storing image and text together as embedding a good option?
When I fire question to elastic, how can i show image and text together to the user?
What are other approaches available for me?

2

Is storing image and text together as embedding a good option?

Yes. I think so. But exclude the embeddings for both fields from the source.

When I fire question to elastic, how can i show image and text together to the user?

You need to add a link (an URL) to the image within you documents. Something like:

{
  "text": "text here",
  "text_emb": [0, 0, 0, ...],
  "image": "https://link-to-img/"
  "image_emb": [0, 0, 0, ...]
}

HTH

3

Thank you for quick response.

Is there any limit on embedding a content ? Any recommendation here?
Each help topic in our case, usually contains10k words and 10-15 images and we have 30-35 such topics.

4

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