Flaglab/banrep-jel-abstracts-es

BanRep JEL Abstracts (Spanish)

This dataset was built by collecting academic and institutional publications from the Banco de la República de Colombia (the Central Bank of Colombia).
Each entry corresponds to a research paper abstract written in Spanish, labeled with one or more JEL (Journal of Economic Literature) classification codes.

The dataset enables multilabel text classification and domain-specific evaluation in the field of economics.

---

Dataset Structure

Split	# Samples
Train	4,463
Validation	900
Test	900

Each record contains the following fields:

Column	Description
input_text	Abstract of the research paper in Spanish.
labels	List of JEL codes assigned to the document.

Example:

{
  "input_text": "Informe de la Evolución de la Balanza de Pagos y de la Posición de Inversión Internacional...",
  "labels": ["E", "P"]
}

---

Objective and Use Cases

The goal of this dataset is to predict the JEL classification codes of research abstracts.
This makes it suitable for tasks such as:

• Multilabel text classification
• Domain-specific language model fine-tuning
• Topic modeling and representation learning in economics
• Evaluation of Spanish encoders (e.g., Sci-BETO, BETO, BERTIN, XLM-R)

---

Data Source and Preprocessing

The dataset was derived from the institutional research repository of the Banco de la República.
All documents were publicly available and written in Spanish.

Preprocessing steps:

1. Extracted abstracts and metadata (title, year, department).
2. Filtered to include only documents with valid JEL codes.
3. Converted multi-label codes (e.g., E, P) into Python lists for supervised training.
4. Translate with GPT 4o the abstracts and titles that were in english to spanish.
5. Split into train, validation, and test sets using a stratified strategy based on JEL labels.

---

Label Space

The dataset includes standard JEL code categories, which cover major areas of economics such as:

Code	Field (Short Description)
A	General Economics and Teaching
B	History of Economic Thought
C	Mathematical and Quantitative Methods
D	Microeconomics
E	Macroeconomics and Monetary Economics
F	International Economics
G	Financial Economics
H	Public Economics
I	Health, Education, and Welfare
J	Labor and Demographic Economics
K	Law and Economics
L	Industrial Organization
M	Business Administration and Marketing
N	Economic History
O	Economic Development, Innovation, Technological Change
P	Economic Systems
Q	Agricultural and Environmental Economics
R	Urban, Rural, and Regional Economics

---

Example Applications

• Fine-tuning RoBERTa or BETO models for multilabel classification.
• Evaluating Spanish encoders on economic abstracts.
• Creating embeddings for clustering by research topic or policy area.
• Benchmarking models on long-text multilabel tasks in the social sciences.

---

Benchmark Context

This dataset is part of the Spanish Scientific NLP Datasets collection, alongside:

• FlagLab/academic-knowledge-abstracts-es
• FlagLab/econ-ie-spanish

All datasets were designed to facilitate evaluation and adaptation of domain-specific Spanish encoders specific for Sci-BETO.

---

License

CC-BY-4.0 — The dataset is derived from publicly accessible documents published by the Banco de la República de Colombia.
Users must provide attribution when using or redistributing this dataset.

---