Using Language Models to make sense of Chat Data without compromising user privacy

submited by

Style Pass

2025-01-05 07:30:04

If you're interested in the code for this article, you can find it here where I've implemented a simplified version of CLIO without the PII classifier and most of the original prompts ( to some degree ).

An ideal solution allows us to understand broad general trends and patterns in user behaviour while preserving user privacy. In this article, we'll explore an approach that addresses this challenge - Claude Language Insights and Observability ( CLIO ) which was recently discussed in a research paper released by Anthropic.

You can read about CLIO in Anthropic's blog post here. Personally I just asked Claude to explain it to me and it did a good job.

We can see this from the diagram below where we have initial private user conversations that get increasingly summarised and clustered so that the broad general trends are preserved while omitting the specific user details.

CLIO was validated by using a largely synthetic dataset that was generated by Claude. This comprised approximately ~20k conversations that spanned across 15 different languages and 18 high level topics.