TheFinAI/plutus-fns2023

---

Dataset Card for FNS-2023 Dataset

Dataset Summary

The FNS-2023 dataset contains Greek text data curated for summarization tasks. This resource is designed to benchmark and advance summarization techniques for Greek language texts, particularly in the financial or general domain where concise summaries are valuable.

Supported Tasks

• Task: Summarization
• Evaluation Metrics: ROUGE-1

Languages

• Greek

Dataset Structure

Data Instances

Each instance in the FNS-2023 dataset comprises three fields:

• query: Typically represents the input text that requires summarization.
• answer: Contains the target summary corresponding to the input text.
• text: Provides the full original context or additional details that the summary is derived from.

Data Fields

• query: String – The primary text or prompt used as input for summarization.
• answer: String – The concise summary produced from the input text.
• text: String – Additional context or the full text from which the summary is generated.

Data Splits

The dataset is distributed into three splits:

• Train: 169 examples (128,656,817 bytes)
• Validation: 43 examples (36,160,573 bytes)
• Test: 50 examples (46,405,723 bytes)

Dataset Creation

Curation Rationale

The FNS-2023 dataset was curated to support research and development on summarization techniques for Greek language texts. By providing diverse examples with a focus on summarization, the dataset seeks to address challenges in reducing lengthy texts to concise summaries without losing critical information.

Source Data

Initial Data Collection and Normalization

• The original authors provided Greek text data comprising training and validation splits.
• Their original validation data was retained and repurposed as the “test” split in this dataset.
• An 80-20 train-validation split was then applied on their original training data to form the current “train” and “validation” splits.

Who are the Source Language Producers?

• The source data was originally curated by the dataset authors as detailed in their GitHub repository:
  https://github.com/iit-Demokritos/FNS2023_data

Annotations

Annotation Process

• The dataset was prepared by collecting and curating Greek text examples and their corresponding summaries.
• No further manual annotations were applied beyond the original curation process.

Who are the Annotators?

• The dataset stems from the original work of the authors who shared the data publicly. No external annotation team was involved beyond this.

Personal and Sensitive Information

• The FNS-2023 dataset does not contain any personally identifiable information (PII) and is strictly focused on Greek text data for summarization purposes.

Considerations for Using the Data

Social Impact of Dataset

By enabling advances in summarization techniques for Greek texts, this dataset can contribute to better information digestion, efficient content processing, and enhanced decision-making processes in both academic and industry settings.

Discussion of Biases

• The dataset is sourced from specific Greek texts and may reflect domain or stylistic biases inherent in the original data.
• Summaries may lean towards the style and linguistic patterns present in the original dataset, which might not generalize across all types of Greek texts.

Other Known Limitations

• Given the relatively small number of examples, the dataset might require augmentation or domain adaptation for more robust model training.
• Variability in text length and summary quality may require additional processing in certain applications.

Additional Information

Dataset Curators

• Original Dataset: Elias Zavitsanos; Aris Kosmopoulos; George Giannakopoulos; Marina Litvak; Blanca Carbajo-Coronado; Antonio Moreno-Sandoval

• Adapted Version:

  • 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: CC BY 4.0

Citation Information

If you use this dataset in your research, please consider citing the following paper:

Original Dataset:

@inproceedings{zavitsanos2023financial,
  title={The financial narrative summarisation shared task (FNS 2023)},
  author={Zavitsanos, Elias and Kosmopoulos, Aris and Giannakopoulos, George and Litvak, Marina and Carbajo-Coronado, Blanca and Moreno-Sandoval, Antonio and El-Haj, Mo},
  booktitle={2023 IEEE International Conference on Big Data (BigData)},
  pages={2890--2896},
  year={2023},
  organization={IEEE}
}

Adapted Version (Plutus):

@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}, 
}