hsilvosa/congreso-debates

Spanish Congress of Deputies: Debates, Verbatim Speeches, and Voting Records (L1 to L15)

High-fidelity dataset containing the parliamentary archive of the Spanish Congress of Deputies (Congreso de los Diputados de España) from the Constituent Legislature / L1 (1979) to the present day (L15).

It includes 41,125 parliamentary interventions with over 188.6 million characters of verbatim speech text extracted across all 1,028 official Daily of Sessions (Diario de Sesiones) PDFs, alongside 105,000 nominal deputy voting records crossed with seat assignments and parliamentary groups.

---

Dataset Summary

Configuration	Rows	Historical Range	Key Metrics	Parquet Size
intervenciones_discursos	41,125 speeches	L1 to L15 (1979–2026)	188,679,518 characters of verbatim spoken text (speech_text), 1,191 unique speakers	7.63 MB
deputy_votes	105,000 votes	L1 to L15 (15 Legislatures)	Individual voting records (Yes, No, Abstain, Did not vote) by deputy, seat, and group	23.8 KB (all) / 15 subsets
initiatives	300 initiatives	L1 to L15	Legislative dossiers, official titles, and plenary vote counts	7.8 KB (all) / 15 subsets

---

Data Structure and Schema

1. intervenciones_discursos (congreso_intervenciones.parquet)

Contains parliamentary interventions with inlined verbatim spoken transcripts:

{
  "legislature": "Leg.15",
  "sesion_date": "29/02/2024",
  "organo": "Pleno",
  "super_tipo_iniciativa": "Tratados y convenios internacionales",
  "tipo_iniciativa": "Convenio internacional",
  "objeto_iniciativa": "Denuncia del Tratado de la Carta de la Energía, hecho en Lisboa el 17 de diciembre de 1994.",
  "num_expediente": "110/000001",
  "autor": "Gobierno",
  "fase": "Debate de totalidad",
  "tipo_intervencion": "Intervención",
  "orador": "Requena Ruiz, Juan Diego",
  "cargo_orador": "Diputado",
  "inicio_intervencion": "10:15:00",
  "fin_intervencion": "10:27:00",
  "speech_text": "DEBATES DE TOTALIDAD DE CONVENIOS INTERNACIONALES...\nLa señora PRESIDENTA: Pasamos al siguiente punto del orden del día...",
  "speech_char_count": 4820,
  "enlace_texto_integro": "https://www.congreso.es/...",
  "enlace_pdf": "https://www.congreso.es/public_oficiales/L15/CONG/DS/PL/DSCD-15-PL-25.PDF#page=12"
}

2. deputy_votes (congreso_deputy_votes_all.parquet)

Individual voting records cast in plenary sessions:

{
  "legislature": 15,
  "session_number": 193,
  "vote_number": 2,
  "date": "23/07/2026",
  "title": "Proposición de Ley relativa a la reforma del régimen fiscal...",
  "deputy_name": "Sánchez Pérez-Castejón, Pedro",
  "parliamentary_group": "GS",
  "seat": "1101",
  "vote": "Sí"
}

3. initiatives (congreso_initiatives_all.parquet)

Metadata and plenary voting results for legislative initiatives:

{
  "legislature": 15,
  "session_number": 193,
  "vote_number": 2,
  "date": "23/07/2026",
  "title": "Proposición de Ley relativa a la reforma del régimen fiscal...",
  "expediente": "122/000045",
  "subgroup_title": "",
  "subgroup_text": "",
  "asentimiento": "No",
  "presentes": 350,
  "afavor": 178,
  "en_contra": 170,
  "abstenciones": 2,
  "no_votan": 0
}

---

Usage

With Python (pandas / polars / pyarrow):

import pandas as pd

# Load verbatim speeches and debates
df_speeches = pd.read_parquet("data/intervenciones/congreso_intervenciones.parquet")

# Filter plenary debates
plenary_debates = df_speeches[df_speeches["organo"] == "Pleno"]
print(f"Total plenary debates: {len(plenary_debates)}")

# Load nominal voting records
df_votes = pd.read_parquet("data/deputy_votes/congreso_deputy_votes_all.parquet")

# Calculate voting distribution across parliamentary groups
voting_dist = df_votes.groupby(["parliamentary_group", "vote"]).size().unstack(fill_value=0)
print(voting_dist)

With Hugging Face datasets:

from datasets import load_dataset

# Load verbatim speeches
dataset_speeches = load_dataset("hsilvosa/congreso-debates", "intervenciones_discursos", split="train")

# Load voting records
dataset_votes = load_dataset("hsilvosa/congreso-debates", "deputy_votes", split="train")

---

Machine Learning and NLP Use Cases

• Spanish LLM Fine-Tuning and Evaluation: Training and evaluating language models on formal rhetoric, parliamentary debate, and legal reasoning in Spanish.
• Stance Detection: Predicting plenary voting behavior (Yes / No / Abstain) directly from speech transcripts and debate text.
• Legal and Political RAG Systems: Semantic search over historical interventions with exact page citations to official proceedings.
• Quantitative Political Science: Quantifying party cohesion, voting discipline, and ideological polarization over 45 years of democratic parliament.

---

Data Collection and Cleaning Methodology

1. Ingestion: Automated query of the official Open Data API from the Congress of Deputies.
2. Document Retrieval: Download of 1,028 official Daily of Sessions (Diario de Sesiones) PDFs.
3. Multi-process Extraction:
   • Extraction of target speech pages using multi-process parallelization across CPU cores.
   • Text normalization with regular expressions removing official document headers, running banners, and pagination stamps.
   • Serialization to typed Apache Parquet with Zstandard compression.

---

Provenance and License

• Source: Congreso de los Diputados Open Data Portal and Official Diarios de Sesiones (Cortes Generales).
• License: Public sector information re-use (Spanish Law 37/2007, EU Directive 2019/1024). Distributed under Creative Commons Attribution 4.0 International (CC-BY 4.0).