README.md

metadata

dataset_info:
  features:
    - name: query
      dtype: string
    - name: answer
      dtype: string
    - name: label
      sequence: string
    - name: text
      dtype: string
  splits:
    - name: train
      num_bytes: 649136
      num_examples: 320
    - name: validation
      num_bytes: 157953
      num_examples: 80
    - name: test
      num_bytes: 230512
      num_examples: 100
  download_size: 271347
  dataset_size: 1037601
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*
license: apache-2.0
task_categories:
  - token-classification
language:
  - gr
tags:
  - finance
  - text
pretty_name: Plutus Finner Text
size_categories:
  - n<1K

---

Dataset Card for Plutus Finner Text

Table of Contents

• Table of Contents
• Dataset Description
  • Dataset Summary
  • Supported Tasks
  • Languages
• Dataset Structure
  • Data Instances
  • Data Fields
  • Data Splits
• Dataset Creation
  • Curation Rationale
  • Source Data
  • Annotations
  • Personal and Sensitive Information
• Considerations for Using the Data
  • Social Impact of Dataset
  • Discussion of Biases
  • Other Known Limitations
• Additional Information
  • Dataset Curators
  • Licensing Information
  • Citation Information
  • Contributions

Dataset Description

• Homepage: https://huggingface.co./collections/TheFinAI/plutus-benchmarking-greek-financial-llms-67bc718fb8d897c65f1e87db
• Repository: https://huggingface.co./datasets/TheFinAI/plutus-finner-text
• Paper: https://arxiv.org/pdf/2502.18772
• Leaderboard: https://huggingface.co./spaces/TheFinAI/Open-Greek-Financial-LLM-Leaderboard#/
• Model: https://huggingface.co./spaces/TheFinAI/plutus-8B-instruct

Dataset Summary

Plutus Finner Text is a dataset crafted for text named entity recognition (NER) within financial documents. Focusing on Greek language financial texts, this resource combines financial queries with answers, labels, and additional contextual text. The dataset is designed as a benchmark to enhance NER capabilities for extracting and categorizing textual entities in finance.

Supported Tasks

• Task: Text Named Entity Recognition
• Evaluation Metrics: Entity F1 Score

Languages

• Greek

Dataset Structure

Data Instances

Each instance in the dataset is composed of four fields:

• query: A financial query or prompt that includes text potentially containing named entities.
• answer: The expected answer associated with the query.
• label: A sequence field containing labels which denote the named entities.
• text: Additional context or commentary that clarifies the query.

Data Fields

• query: String – Represents the financial query or prompt.
• answer: String – The corresponding answer for the query.
• label: Sequence of strings – Contains the named entity labels linked to each instance.
• text: String – Provides supplementary context or details.

Data Splits

The dataset is organized into three splits:

• Train: 320 instances (649,136 bytes)
• Validation: 80 instances (157,953 bytes)
• Test: 100 instances (230,512 bytes)

Dataset Creation

Curation Rationale

The Plutus Finner Text dataset was developed to support robust text-based named entity recognition in the financial domain, tailored specifically for Greek language texts. It aims to empower researchers and practitioners with a challenging benchmark for extracting and classifying named entities within financial documents.

Source Data

Initial Data Collection and Normalization

The source data was derived from a diverse collection of Greek financial annual reports containing numeric information.

Who are the Source Language Producers?

Greek financial annual reports.

Annotations

Annotation Process

The annotation process involved domain experts in both finance and linguistics who manually identified and marked the relevant named entities within the financial queries and contextual text. Quality control was maintained to ensure high annotation consistency.

Who are the Annotators?

A collaboration between financial analysts, data scientists, and linguists was established to annotate the dataset accurately and reliably.

Personal and Sensitive Information

This dataset has been curated to exclude any personally identifiable information (PII) and focuses solely on financial textual data and entity extraction.

Considerations for Using the Data

Social Impact of Dataset

By advancing text NER within the Greek financial sector, this dataset supports improved information extraction and automated analysis—benefiting financial decision-making and research across the industry and academia.

Discussion of Biases

• The domain-specific language and textual formats may limit generalizability outside Greek financial texts.
• Annotation subjectivity could introduce biases in the identification of entities.
• The dataset’s focused scope in finance may require further adaptation for use in broader contexts.

Other Known Limitations

• Additional pre-processing might be needed to handle variations in text and entity presentation.
• The dataset’s application is primarily limited to the financial domain.

Additional Information

Dataset Curators

• Xueqing Peng
• Triantafillos Papadopoulos
• Efstathia Soufleri
• Polydoros Giannouris
• Ruoyu Xiang
• Yan Wang
• Lingfei Qian
• Jimin Huang
• Qianqian Xie
• Sophia Ananiadou

The research is supported by NaCTeM, Archimedes RC, and The Fin AI.

Licensing Information

• License: Apache License 2.0

Citation Information

If you use this dataset in your research, please consider citing it as follows:

@misc{peng2025plutusbenchmarkinglargelanguage,
      title={Plutus: Benchmarking Large Language Models in Low-Resource Greek Finance}, 
      author={Xueqing Peng and Triantafillos Papadopoulos and Efstathia Soufleri and Polydoros Giannouris and Ruoyu Xiang and Yan Wang and Lingfei Qian and Jimin Huang and Qianqian Xie and Sophia Ananiadou},
      year={2025},
      eprint={2502.18772},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2502.18772}, 
}