shevlinn/dail-pq

一个机器标注的语料库，包含2002年至2026年间爱尔兰众议院（Dáil Éireann）提交的每一个书面议会问题（PQ）——约965,000个问题，每个问题都标注了十三个分析维度，包括本地与全国范围、政策与行政、受益人、提问类型、比较议程项目主题和回答实质性等。标签由自托管的Qwen 3 32B模型生成，并通过与Anthropic Claude Haiku 4.5、Qwen 3 27B和Qwen 3 32B的500个问题比较集进行了验证。基础议会记录来自公共Oireachtas API，数据集的贡献在于标签和模式。A machine-labelled corpus of every written parliamentary question (PQ) submitted in Dáil Éireann between 2002 and 2026 — roughly 965,000 questions, each annotated for thirteen analytic dimensions including local-vs-national scope, policy-vs-administration, beneficiary, ask type, Comparative Agendas Project topic, and reply substantiveness. The labels were produced by a self-hosted Qwen 3 32B (FP8) instance, with validation against a 500-question comparison set labelled by Anthropic Claude Haiku 4.5, Qwen 3 27B, and Qwen 3 32B documented in the project paper. The underlying parliamentary record is the public Oireachtas API; our contribution is the labels and the schema.