Quantification of Biodiversity from Historical Survey Text with LLM-based Best-Worst Scaling
Thomas Haider, Tobias Perschl, Malte Rehbein · 2025
Read Quantification of Biodiversity from Historical Survey Text with LLM-based Best-Worst Scaling by Thomas Haider, Tobias Perschl, Malte Rehbein (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-2502-04022.txt.
Description
Counting how common species were in the past, using old survey texts, is the task set by Thomas Haider, Tobias Perschl, and Malte Rehbein in this 2025 paper. The texts rarely give exact numbers, so the team tested ways to turn written reports into quantity estimates. They compared plain classification with a ranking method called best-worst scaling.
In best-worst scaling, a model picks the most and least common items in a small set, and scores build up from many such picks. Three large language models were tested against human judgments. Two of them agreed well with humans and with each other. The authors conclude the ranking route is cheaper and about as robust as finer classification.
Catalog record
| CREATOR | Thomas Haider, Tobias Perschl, Malte Rehbein |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-04022