SocCor: A Multimodal-based Multilingual Soccer Corpus for Text Data Analytics

Overview

SocCor is a comprehensive multilingual soccer text corpus developed for text data analytics and research purposes. The corpus contains annotated soccer commentaries and reports in various languages and formats.

Project Structure

SocCor/
├── Textdata/
│   ├── raw_data/
│   │   ├── Reports/
│   │   |   ├── BBC/
│   │   |   ├── Kicker/ 
│   │   |   └── ... 
│   |   ├── Games/
│   │   |   ├── BBC/
│   │   |   ├── Sportschau/ 
│   |   |   └── ...
│   |   └── ...
│   └── cleaned_data/
│       ├── Reports/
│       |   ├── BBC/
│       |   ├── Kicker/ 
│       |   └── ... 
│       ├── Games/
│       |   ├── BBC/
│       |   ├── Sportschau/ 
│       |   └── ...
│       └── ...
└── Metadata/   
     ├── Players
     └──  ...

Data Sources

The corpus includes various types of soccer text data:

Reports

• Format: JSON
• Content: Soccer match summaries with annotated player information

Games

• Format: JSON with timestamps
• Content: Live match commentaries (transcribed)

Highlights

• Format: JSON with timestamps
• Content: Key moments and turning points of matches (transcribed)

Livetickers

• Format: JSON with timestamps
• Content: Real-time match updates in written format

Data Annotation

Player Annotation

Player names are annotated with structured tags:

<PLAYERNAME_POSITION_COUNTRY>

Examples:

• <BELLINGHAM_ATTACKING-MIDFIELD_ENG> - Jude Bellingham, Attacking Midfielder, England
• <PICKFORD_GOALKEEPER_ENG> - Jordan Pickford, Goalkeeper, England

Positions

• GOALKEEPER - Goalkeeper
• CENTRE-BACK - Centre-back
• LEFT-BACK / RIGHT-BACK - Full-backs
• DEFENSIVE-MIDFIELD - Defensive midfielder
• CENTRAL-MIDFIELD - Central midfielder
• ATTACKING-MIDFIELD - Attacking midfielder
• LEFT-WINGER / RIGHT-WINGER - Wingers
• CENTRE-FORWARD - Centre-forward

Country Codes

• ENG - England
• TUR - Turkey
• CZE - Czech Republic
• ...

Data Format

JSON Structure (Games & Highlights)

{
    "start": 0.0,
    "end": 5.0,
    "text": "Transcribed live commentary with <PLAYER_POSITION_COUNTRY> annotations"
}

JSON Structure (Livetickers)

{
    "minute": 12,
    "text": "Real-time match updates in written format with <PLAYER_POSITION_COUNTRY> annotations"
}

JSON Structure (Reports)

{
    "text": "Soccer match summaries with <PLAYER_POSITION_COUNTRY> annotations"
}

Usage

The corpus is suitable for various NLP tasks:

• Named Entity Recognition (NER) - Recognition of players, clubs, positions
• Sentiment Analysis - Evaluation of player performances
• Event Extraction - Identification of match events
• Multilingual Text Analysis - Comparison of different language styles
• Time Series Analysis - Timeline of match commentaries

Features

Multimodal Annotations

• Player names with position and nationality
• Temporal annotations for live commentary

Citation

If you use this corpus in your research, please cite:

@inproceedings{lohr2025soccor,
    title     = {SocCor: A Multimodal-based Multilingual Soccer Corpus for Text Data Analytics},
    author    = {Paul L{\"o}hr and Jannik Str{\"o}tgen},
    booktitle = {KONVENS 2025},
    year      = {2025},
    url       = {https://openreview.net/forum?id=iYQGEuI47t}
}

Contact

For questions or comments regarding the SocCor corpus, please contact:

paul.loehr@h-ka.de

---

This project is part of research in sports text analytics and multilingual NLP applications.