stefan-it/co-funer

---
license: mit
task_categories:
- token-classification
language:
- de
---

# CO-Fun: A German Dataset on Company Outsourcing in Fund Prospectuses for Named Entity Recognition and Relation Extraction

This inofficial dataset repository provides a CoNLL-like version of the CO-Fun **NER** dataset, that was proposed in the CO-Fun paper (https://arxiv.org/abs/2403.15322):

> The process of cyber mapping gives insights in relationships among financial entities and service providers. Centered around the outsourcing practices of companies within fund prospectuses in Germany, we introduce a dataset specifically designed for named entity recognition and relation extraction tasks. The labeling process on 948 sentences was carried out by three experts which yields to 5,969 annotations for four entity types (Outsourcing, Company, Location and Software) and 4,102 relation annotations (Outsourcing-Company, Company-Location). State-of-the-art deep learning models were trained to recognize entities and extract relations showing first promising results.

## Preprocessing

The notebook [Export-To-CoNLL.ipynb](Export-To-CoNLL.ipynb) performs the necessary steps to create a CoNLL-like version of the CO-Fun dataset, that could easily be used for fine-tuning NER models.

Additionally, the [FlairDatasetTest.ipynb](FlairDatasetTest.ipynb) notebooks loads the dataset with the Flair dataset loader and checks, if the number of parsed sentences is correct and identical to the number of sentences reported in the official CO-Fun paper.

## Named Entites

The CO-Fun dataset provides annotations for the following Named Entities:

* `Auslagerung` (engl. outsourcing)
* `Unternehmen` (engl. company)
* `Ort` (engl. location)
* `Software`

# Example: Load Dataset with Flair library

The notebooks [FlairDatasetExample.ipynb](FlairDatasetExample.ipynb) shows how to load the dataset with the awesome [Flair library](https://github.com/flairNLP/flair).

# Changelog

* 25.03.2024: Initial version of the preprocessed CO-Fun NER dataset is released.

# Licence

The original CO-Fun dataset is released under MIT license. Thus, this preprocessed version is also licenced under MIT.

原始信息汇总

CO-Fun: A German Dataset on Company Outsourcing in Fund Prospectuses for Named Entity Recognition and Relation Extraction

Overview

• License: MIT
• Task Categories: Token-classification
• Language: German (de)

Dataset Description

• Purpose: Designed for named entity recognition (NER) and relation extraction tasks.
• Content: Annotated 948 sentences with 5,969 annotations for four entity types (Outsourcing, Company, Location, and Software) and 4,102 relation annotations (Outsourcing-Company, Company-Location).
• Expertise: Labeling process conducted by three experts.
• Results: Trained state-of-the-art deep learning models showing promising results in entity recognition and relation extraction.

Named Entities

• Auslagerung (Outsourcing)
• Unternehmen (Company)
• Ort (Location)
• Software

Preprocessing

• Tools: Export-To-CoNLL.ipynb for creating a CoNLL-like version of the dataset.
• Verification: FlairDatasetTest.ipynb for checking the number of parsed sentences against the official CO-Fun paper.

Usage Example

• Library: Flair
• Example: FlairDatasetExample.ipynb demonstrates loading the dataset using the Flair library.