Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

After All Is Said and Indexed – Unlocking Information in Recorded Speech

github.com

Post reply on HN
Print view View original

11–17 of 17 posts

Re: After All Is Said and Indexed – Unlocking Information in Recorded Speech

#11

Post by jeadie »

notjulianjaynes wrote: post #9

This is interesting but what problem does it solve better than CTRL+F-ing a transcript? It seems like this would be a worse solution for when the precise way someone says something could be important (ex. journalists parsing an interview, students studying their recorded lectures) and that it would be most useful if you were working with a large volume of recorded audio, such as customer service calls. This makes me…

Being able to handle and ask questions of audio data is a pretty big field. https://www.assemblyai.com/, for example, is a company entirely dedicated to audio intelligence. They have some great example use cases on their page.

Re: After All Is Said and Indexed – Unlocking Information in Recorded Speech

#12

Post by jeadie »

rektide wrote: post #4

Hadn't heard of the thing they were putting their data into, Marqo, a "tensor search for humans" , https://github.com/marqo-ai/marqo

Its a great tool. Unlike vectorDBs alone, Marqo helps the full process that alot of people end up wanting to use vectorDBs for (e.g. have structured data, use LLMs to create embeddings, and perform search/CRUD on embeddings + original data).

Re: After All Is Said and Indexed – Unlocking Information in Recorded Speech

#13

Post by jeadie »

moneywoes wrote: post #8

How does this compare to using Whisper and feeding that into a vector DB and querying with a LLM Pardon the dumb question I only have an elementary understanding

Not a dumb question at all! Essentially what can do Marqo, and this blog shows, is that there is alot of logic and work to do what you said (i.e. pass raw data into LLM, get embeddings, store in vector DB, then query both embeddings and original data).

Re: After All Is Said and Indexed – Unlocking Information in Recorded Speech

#14

Post by jeadie »

thomasahle wrote: post #6
jeadie wrote: post #2

A really interesting blog post I found using LLMs for audio search which I think is a pretty nifty/new idea. I've found it cumbersome using some of the new vector DBs (chroma, faiss, etc) to make end to end systems, but with Marqo it doesn't seem too hard.

> I've found it cumbersome using some of the new vector DBs (chroma, faiss, etc) to make end to end systems What parts are cumbersome?

Most people, like me, who end up needing to use vector DBs, are wanting to use LLMs on a specific, often private dataset/use case. Typically one starts with something like unstructured JSON data, then need to pick and manage LLMs to create embeddings, then store these and the original JSON data in a vectorDB. Then the application is some variety of CRUD operations + searching over both the original data and the embeddings.

Chroma, Pinecone, I guess FAISS/HNSWlib/etc only handle vector operations. Really what I'd want, which Marqo does, is handle everything end to end.

Post reply on HN