TBMM Plan and Budget Committee Discourse Corpus

A structured, machine-readable corpus of the Turkish Grand National Assembly (TBMM) Plan and Budget Committee (Plan ve Bütçe Komisyonu, PBK) budget proceedings. In Türkiye, the Plan and Budget Committee is the first and most detailed parliamentary stage where the central government budget is deliberated before plenary debate. Each ministry's budget is discussed in depth; ministers, bureaucrats, and members of parliament engage in extensive technical exchanges. Committee sessions are typically held in October-November each year. ==============================CURRENT VERSION COVERAGE============================== This Zenodo concept record points to all versions of the corpus. The specific coverage and contents of each version are documented in CHANGELOG.md within the repository, and in each version's individual description. The initial release (v1.0) covers:- Budget years: 2009-2025 (17 years)- Speaker turns: 223,408- Source documents: 294 PDFs- Linked MPs: 1,184 unique members- Ministers: ~100 (both MP-ministers and appointed technocrats)- Speaker metadata linkage rate: 97.5% Future versions will extend coverage as new budget year proceedings become available, and may include retrospective extensions to earlier periods, refinements to existing metadata, and structural improvements documented in the changelog.Update note (2026-05-30): Data files (processed parquet and raw PDFs) added to this record. The metadata, version, and DOI remain unchanged from the original v1.0.1 publication. ==============================DATA STRUCTURE============================== The corpus is provided as a long-format Parquet file where each row represents a single speaker turn, with linked metadata:- konusma_id: unique speech identifier- tarih: speech date- butce_yili: budget year being deliberated- konusmaci_ham / konusmaci_sade: raw and normalized speaker name- rol: speaker role (milletvekili / bakan / baskan / burokrat)- parti: matched party affiliation- sehir: province- sicil: TBMM permanent MP identifier- tbmm_donem: TBMM term number- metin: speech text- kelime_sayisi: word count ==============================SOURCES============================== Two official sources are integrated:- Strategy and Budget Department (SBB) archive for 2016-2025 (196 PDFs)- TBMM legacy transcript system for 2009-2015 (98 PDFs) ==============================METHODOLOGY HIGHLIGHTS============================== - Four-tier MP-speech matching algorithm using TBMM permanent identifiers (sicil numbers)- Manual corrections documented for MPs missing from official rosters- Coverage verification via meeting number sequencing and ministry presence checks- Documented encoding fixes (118,596 character corrections in 2016 PDFs)- Poppler-based text extraction for PDF 1.6 non-linearized documents (2018 batch) Full methodology is documented in the GitHub repository. ==============================POTENTIAL USE CASES============================== - Time-series analysis of fiscal discourse- Minister-level and party-level rhetoric comparisons- Crisis period parliamentary response analysis- Discourse-policy lag analysis- Methodological work on Turkish parliamentary text-as-data- Comparative legislative studies (with related corpora below) ==============================RELATED CORPORA============================== This corpus is committee-focused and complements existing Turkish parliamentary resources: - Demirtas, E. (2026). TBMM Parliamentary Proceedings Corpus, 1950-2023 — plenary proceedings. https://doi.org/10.5281/zenodo.19713325- Erjavec, T. et al. (2025). ParlaMint 5.0 — multilingual parliamentary debates including Turkish plenary 2011-2021. ==============================DOCUMENTATION============================== For full methodology, data dictionary, replication guide, known issues, and coverage verification, see the GitHub repository:https://github.com/eozyerden/tbmm-pbk-corpus ==============================LICENSES============================== - Code (R scripts): MIT License- Data and documentation: CC BY 4.0 ==============================HOW TO CITE============================== When citing this corpus in academic work, please cite the specific version you used. Each version has its own DOI. The concept DOI (this page) always points to the latest version. For specific version citation, see the "Cite" widget on each version's page.