Exploring Embeddings: Workflow Tips & Real Use-Cases!

The Best Of

Feb 7, 2024, 12:59 AM

I’m curious: Anyone using embeddings? What’s your workflow? I’m still toying around with mapping the data but I am looking for “real” use-cases

topic mapping chrome extension prototype

Forwarded thread from another channel:

Dale McGeorge

Feb 7, 2024, 1:02 AM

Im using embeddings a lot. I'm currently working on a task that compares different embedding models for keyword classification. I hope to present it with Noah and Joy on the webinar coming up at the end of the month.

Dale McGeorge

Feb 7, 2024, 1:03 AM

The workflow will be use case dependent, so ours is to embed data as we push it into the database so that our RAG process can use it in a prompt

Valentin Pletzer

Feb 7, 2024, 1:06 AM

Sounds interesting. You create embeddings of the keywords themselves or is it more an embedding created of a piece of text representing the keyword (like a snippet or description)?

Dale McGeorge

Feb 7, 2024, 1:07 AM

We embed lots of things, like Google Reviews and keywords then depending on the task we can call upon certain types of data.

ah ok. got it

for me it’s mostly title + description

Dale McGeorge

Feb 7, 2024, 1:10 AM

Have you found a good 2D plotting algorithm for vectors?

Valentin Pletzer

Feb 7, 2024, 1:12 AM

I’m using umap for dimension reduction and dbscan for clustering

Feb 7, 2024, 7:20 AM

@pletzer - I’ve been mucking around with embeddings as well to identify internal links ()

Valentin Pletzer

Feb 7, 2024, 7:31 AM

love the article. Cool idea and really easy to read (and probably replicate)

Feb 7, 2024, 7:43 AM

@pletzer happy to hear that! Feel free to shoot me a DM if you have any questions

Dale McGeorge

Feb 7, 2024, 4:42 PM

Nice one @jordan.choo, Have you tried any chunking methods? It might also help refine where to put the link within the content.

Feb 8, 2024, 7:57 AM

@dale I have not and didn’t even think about it. I’ll have to give that a whirl

Kristin Tynski

Feb 12, 2024, 10:43 AM

Yes, doing a lot with embeddings for content creation pipelines Im working on. Useful in many ways. Im using it for my RAG/Retrieval, and doing some fun stuff with neo4j and integrating graph traversal with nearest neighbor embedding lookup on a knowledge graph

Mihir Naik

Feb 21, 2024, 8:50 AM

I wrote a script that does sentence level embeddings and help find internal links

The script isn't public yet, but I will be publishing soon.

How this script works: Advertools with body text | Mihir Naik posted on the topic | LinkedIn

Yesterday evening I wrote a script that takes "anchor text" & "destination URL" as an input, and suggest Top 3 Sentences & the URL where that sentence is from… | 13 comments on LinkedIn