The Critical Role of Aspects in Measuring Document Similarity

Eftekhar Hossain, Tarnika Hazra, Ahatesham Bhuiyan, Santu Karmaker · 2026

Read The Critical Role of Aspects in Measuring Document Similarity by Eftekhar Hossain, Tarnika Hazra, Ahatesham Bhuiyan, Santu Karmaker (2026) 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-2601-03435.txt.

Description

Eftekhar Hossain and three co-authors say document similarity should always name the aspect being compared. Two texts might match on topic but differ on method, stance, or time period, so one overall score can mislead. Their framework, ASPECTSIM, asks raters and models to judge similarity for a stated aspect only.

The team built a benchmark of about twenty-six thousand aspect and document pairs. With direct GPT-4o prompting, ASPECTSIM matched human judgments far better than holistic scoring did. Smaller open models did poorly when prompted directly, though a two-stage fix raised their agreement a lot. Even then they lagged behind the large model, a warning for anyone building cheap similarity tools.

Catalog record

CREATOREftekhar Hossain, Tarnika Hazra, Ahatesham Bhuiyan, Santu Karmaker
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE95 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-03435

Browse the full library · About the Tower of Babel