From 124 Million Tokens to 1,021 Neologisms: A Large-Scale Pipeline for Automatic Neologism Detection

Diego Rossini, Lonneke van der Plas


Abstract
We present a scalable, modular pipeline for automatic neologism detection that combines rule-based filtering with LLM classification. The pipeline is grounded in two complementary word-formation frameworks, grammatical and extra-grammatical morphology, which jointly define the scope of what counts as a neologism and inform a four-class classification scheme (NEOLOGISM, ENTITY, FOREIGN, NONE). While designed to be modular and transferable at the architectural level, the pipeline is instantiated on 527 million English-language Reddit posts spanning 2005–2024. From this corpus, we extract 124.6 million unique tokens and reduce them by over 99.99% to yield 1,021 neologism candidates, a set small enough for manual expert verification. Multiple LLMs independently classify each candidate via majority vote, with a final verification step, revealing substantial cross-model disagreement and highlighting the challenge of operationalizing neologism detection at scale. Manual annotation of all 1,021 candidates confirms that 599 (58.7%) are genuine lexical innovations.
Anthology ID:
2026.neollm-1.1
Volume:
Proceedings of the Workshop Neology and Large Language Models
Month:
May
Year:
2026
Address:
Palma, Mallorca (Spain)
Editors:
Giedre Valunaite Oleskeviciene, Voula Giouli, Florentina Armaselu, Chaya Liebeskind, Barbara McGillivray
Venues:
NeoLLM | WS
SIG:
Publisher:
ELRA Language Resources Association (ELRA)
Note:
Pages:
1–15
Language:
External URL:
https://lrec.elra.info/lrec2026-ws-neollm-01
DOI:
10.63317/4o6ks86o293r
Bibkey:
Cite (ACL):
Diego Rossini and Lonneke van der Plas. 2026. From 124 Million Tokens to 1,021 Neologisms: A Large-Scale Pipeline for Automatic Neologism Detection. In Proceedings of the Workshop Neology and Large Language Models, pages 1–15, Palma, Mallorca (Spain). ELRA Language Resources Association (ELRA).
Cite (Informal):
From 124 Million Tokens to 1,021 Neologisms: A Large-Scale Pipeline for Automatic Neologism Detection (Rossini & van der Plas, NeoLLM 2026)
Copy Citation: