Tensor Topic Modeling Via HOSVD

Yating Liu, Claire Donnat · 2025

Read Tensor Topic Modeling Via HOSVD by Yating Liu, Claire Donnat (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-2501-00535.txt.

Description

Yating Liu and Claire Donnat extend topic modeling to data with extra dimensions, in a paper whose head dates it to 2024. Topic models sort documents into themes, treating each document as a mix of topics. Fast recent methods handle plain document piles well, but they cannot easily use time, place, or other labels attached to each text.

Their method reshapes the data as a tensor, a grid with several axes, and applies a higher-order splitting method to estimate topics and clusters at once. A core grid then shows how the hidden factors interact. They prove how fast the estimates converge, in a strict entry-by-entry sense. Tests on made-up data and on research abstracts showed the method recovers patterns across dimensions better than flat approaches.

Catalog record

CREATORYating Liu, Claire Donnat
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE252 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-00535

Browse the full library · About the Tower of Babel