Skip to content
Trending storyDeveloping

Google DeepMind's CALLIOPE builds private chatbot topic trees from keywords

1 article1 sourcesince Oct 28Last article 1h ago ·

Overview

AISummary of 1 article

Google DeepMind researchers propose CALLIOPE, a privacy-preserving method for analyzing chatbot usage.

An LLM tags each document with keywords, noise is added so only commonly used keywords survive, and topics are then built from those keywords. Rohan Paul, writing on X, reports that at the strictest privacy setting tested, human raters preferred CALLIOPE over URANIA, an embedding-based method, on all three datasets, by up to 76% versus 18%. Paul also reports that CALLIOPE uses 2 to 5 times more tokens.

Written by AI from the articles below · updated Oct 28, 2:42 PM ET

Check the sources:

Latest developmentOct 28, 2:32 PM ET
Google DeepMind's CALLIOPE builds clearer private topic maps from keywords

Article timeline

The articles in this story. Times are ET.

Oct 28
  1. Rohan PaulX
    Google DeepMind's CALLIOPE builds clearer private topic maps from keywords

    AIGoogle DeepMind researchers propose CALLIOPE, which has an LLM tag documents with keywords, adds noise so only common keywords survive, and then builds topics from them. At the strictest privacy setting tested, human raters preferred CALLIOPE over the embedding-based URANIA on all three datasets, by up to 76% versus 18%. The method uses 2 to 5 times more tokens.

    Image from @rohanpaul_ai's post

Heat Trend

Not enough continuous observations to show a trend yet.