Using language models to label clusters of scientific documents

Dakota Murray, Chaoqun Ni, Weiye Gu, Trevor Hubbard · 2025

Read Using language models to label clusters of scientific documents by Dakota Murray, Chaoqun Ni, Weiye Gu, Trevor Hubbard (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-02601.txt.

Description

Dakota Murray, Chaoqun Ni, Weiye Gu, and Trevor Hubbard study a small task with outsized influence on how scientists navigate the literature: naming clusters of papers. Bibliometric software groups documents into topics, and the old way of labelling a cluster, gluing together its most distinctive terms, produces terse, cryptic names. Generative language models can write fluent, human-like labels instead, and databases have adopted them quickly, ahead of any careful study of what makes such a label good.

The authors separate two label types, characteristic labels and descriptive labels, and give descriptive labelling a formal definition with its design decisions spelled out. They propose a structured generation workflow for bibliometric practice and an evaluation framework for judging the outputs. In their tests, model-written descriptive labels performed at or near the level of characteristic labels, and the paper closes with design guidance for using them responsibly.

Catalog record

CREATORDakota Murray, Chaoqun Ni, Weiye Gu, Trevor Hubbard
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE77 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-02601

Browse the full library · About the Tower of Babel