A Dictionary of Hiberno-English

This dataset provides a machine-readable, structured representation of A Dictionary of Hiberno-English (3rd edition, 2020), compiled and edited by Terence Patrick Dolan and published by Gill Books. It contains 3,899 lexical entries, organised into a consistent schema covering headwords, variant forms, pronunciation, parts of speech, senses and definitions, etymology, cross-references, regional metadata, and usage notes. The deposit includes two complementary datasets. HYBRID is the production dataset, generated through a hybrid extraction pipeline combining a deterministic rule-based parser with large language models. Extraction systems were selected independently for different field groups using development-set evaluation. RB provides the complete output of the standalone deterministic rule-based parser, enabling comparison between rule-based and hybrid approaches. Both datasets are provided in JSON and Microsoft Excel formats, accompanied by a shared JSON Schema, human-readable schema documentation, and detailed methodological and accuracy information. Extraction accuracy was evaluated using two disjoint, stratified reference samples of 390 entries each: E390, produced through machine-led adjudication with targeted human review, and H390, an independently constructed, human-validated held-out reference set. The HYBRID dataset achieved an overall targeted-extraction F1 score of 0.836 against H390. Performance varies substantially by field, with usage notes exhibiting particularly low accuracy (F1 = 0.026). The dataset should therefore be treated as an imperfect computational representation of the original dictionary rather than an authoritative replacement for it. Not all entries have been individually verified against the source. The dataset supports computational research in Hiberno-English, lexicography, corpus linguistics, natural language processing, and digital humanities. The dataset is derived from Dolan's original work and is distributed with permission from Gill Books under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International licence (CC BY-NC-SA 4.0). Copyright in the original work remains with its respective rightsholders. The extraction code and associated methodological resources are available at:https://github.com/RasikaED/hiberno-english-grounded-reader Funding: CASCADE, a Horizon Europe Marie Skłodowska-Curie Actions Doctoral Network (Grant Agreement No. 101119511), supported by the European Commission and UKRI. DOI: 10.5281/zenodo.23882722