English

Information-Theoretic Generative Clustering of Documents

Machine Learning 2024-12-19 v1 Computation and Language Information Retrieval Information Theory math.IT

Abstract

We present {\em generative clustering} (GC) for clustering a set of documents, X\mathrm{X}, by using texts Y\mathrm{Y} generated by large language models (LLMs) instead of by clustering the original documents X\mathrm{X}. Because LLMs provide probability distributions, the similarity between two documents can be rigorously defined in an information-theoretic manner by the KL divergence. We also propose a natural, novel clustering algorithm by using importance sampling. We show that GC achieves the state-of-the-art performance, outperforming any previous clustering method often by a large margin. Furthermore, we show an application to generative document retrieval in which documents are indexed via hierarchical clustering and our method improves the retrieval accuracy.

Keywords

Cite

@article{arxiv.2412.13534,
  title  = {Information-Theoretic Generative Clustering of Documents},
  author = {Xin Du and Kumiko Tanaka-Ishii},
  journal= {arXiv preprint arXiv:2412.13534},
  year   = {2024}
}

Comments

Accepted to AAAI 2025

R2 v1 2026-06-28T20:39:55.579Z