Data

Has anyone been doing large scale thematic analysis of free text using LLMs?

I'm about to start doing this at my startup and wondered if anyone had ideas about best practice and prompting 

Please recast to other relevant channels! 🙏

Be specific.     

leonasskau.co.uk //
Hosting /strategy, /leo, /geopolitics // built Culture3.com // online: leosn.xyz
d33m:mun

Building Higher apparel at High Line 🧢 
Host at /clickcast /cpg || Husband to @shiwen

Ie: identifying the themes across lots of free text comments

Thanks

I’ll write the code and maybe even pick the categories
 I’m mainly interested in the prompt architecture

I have! Tried it for work.

heymarvin and a bunch of startups are trying to make this good. can share all of them if you want.

just using Claude/OAI has mixed results, it kinda just makes stuff up that sounds “right”, but it struggles working with nuance unless you prompt it properly.

3 things impact quality, SCOPE, SPECIFICITY, and CONTEXT

You want very tight scope and specificity with a lot of context. 

That might mean asking the LLM to code and create categories line by line.

aichannel

prev founded rocketr.net, swish.id now exploring aiXcrypto identity ai. expired sprint car driver & energy data engineer. idol worship on shoni_eth.X

I will see any insights you share in /aichannel

lemme know if you want to hop on a call to discuss this, i do it a ton (eg for a project with a training hospital we’re taking reviews of the residents and classifying/extracting concepts) // i’ve done it a ton for curating LLM training data too