Re:Ask HN: What Are You Working On? (October 2024)
by tikimcfee on Monday October 28, 2024 @08:39AM
(#41970474)
I'm obsessed with visual space representations of word as well. My application has a test mode where I render an entire dictionary (like, Webster) and then "plays" a book word by word, highlighting the word, then it's definition, and those words definitions and so on and so on, and it creates a trippy and distinct visualization kinda like an audio visualizer.. but with words. We should chat if you're interested in this kinda thing!
https://github.com/tikimcfee/LookAtThat
Re:Ask HN: What Are You Working On? (October 2024)
by nutanc on Monday October 28, 2024 @09:13AM
(#41970724)
This is really cool.
Re:Ask HN: What Are You Working On? (October 2024)
by tatubola on Monday October 28, 2024 @03:01PM
(#41974800)
Using this for semantic chunking feels like a great idea. Did you continue investigating that (after the second newsletter)?
Re:Ask HN: What Are You Working On? (October 2024)
by nutanc on Monday October 28, 2024 @09:46PM
(#41978553)
Yes. We have integrated this into our RAG pipeline. Looking for a benchmark which we can try. Have been using this to give feedback to my marketing team on our blog posts.
Also, as I mentioned, now, trying out experiments in the video domain with some pretty cool results. Will post the details this week.
Started of with [1] which showed that there might be some strength to the idea.
Applied it for chunking [2] and web site analysis[3] and got pretty good results.
Just started trying out experiments on video [4] and surprised that the structure seems to hold for image embeddings as well.
I have no clue if this has any value, but it is fun to go down this rabbit hole :)
[1]https://gpt3experiments.substack.com/p/the-shape-of-stories-...
[2]https://gpt3experiments.substack.com/p/a-new-chunking-approa...
[3]https://gpt3experiments.substack.com/p/using-semantic-chunki...
[4]https://x.com/nutanc/status/1849661242027409723