arnomatic/german-opus-subtitles

German OPUS OpenSubtitles

Dataset Description

This dataset contains German movie and TV subtitles from the OPUS OpenSubtitles corpus. It provides a large collection of natural, conversational German text extracted from movie and TV show subtitles.

Key Features

• 141,565,623 lines of German dialogue
• 4.2 GB of clean text data
• 92.5% unique lines (low duplication rate)
• Natural conversational German across diverse genres
• Minimal contamination (0.2% English, 0.8% ALL CAPS headers)

Dataset Statistics

Metric	Value
Total Lines	141,565,623
Unique Lines	~130M (92.5%)
File Size	4.2 GB
Avg Line Length	~30 chars
Language	German (99.8%)

Quality Metrics

✅ Unique rate: 92.5%
✅ Duplicates: 7.5% (mostly natural repetitions like "Ja.", "Nein.")
✅ ALL CAPS: 0.8% (location headers, minimal)
✅ English: 0.2% (negligible)
✅ Very short: 0.2%
✅ Empty lines: 0%

Dataset Source

• Original Corpus: OPUS OpenSubtitles
• Language: German (de)
• Domain: Movie and TV subtitles
• Collection Method: Extracted from subtitle files

Use Cases

This dataset is ideal for:

• Conversational AI training - Natural dialogue patterns
• Language modeling - Diverse German text covering multiple registers
• Dialogue systems - Learning turn-taking and conversational flow
• Text generation - Realistic German sentence structure
• German NLP research - Large-scale corpus for various tasks

Data Structure

Each line in the dataset represents a single subtitle line:

{
    "text": "Wie geht es dir?"
}

Usage

Loading with HuggingFace Datasets

from datasets import load_dataset

# Streaming (recommended for large datasets)
dataset = load_dataset(
    "arnomatic/german-opus-subtitles",
    split="train",
    streaming=True
)

# Shuffle
dataset = dataset.shuffle(seed=42, buffer_size=10000)

# Iterate
for example in dataset:
    print(example["text"])

Loading without streaming

from datasets import load_dataset

dataset = load_dataset(
    "arnomatic/german-opus-subtitles",
    split="train"
)

print(f"Total examples: {len(dataset)}")

Sample Examples

"Ihr habt meine Jungs verprügelt?"
"Hatte seit Neuestem seine eigene Fernsehsendung."
"Ortis und Green, das berühmteste Klamauk-Duo von Paris."
"Ich bin gerade echt im Arsch."
"Warum musste das passieren?"
"Wer weiß?"
"Ja, ja, so ist das Leben..."

Dataset Characteristics

Duplication Analysis

Most repeated lines (natural for dialogue):

• "Ja." (70 occurrences)
• "Nein." (39 occurrences)
• "Danke." (33 occurrences)
• "Was?" (21 occurrences)
• "Okay." (20 occurrences)

These repetitions reflect natural conversational patterns rather than data quality issues.

Content Distribution

The dataset covers diverse genres:

• Action & Thriller
• Drama & Romance
• Comedy & Sitcoms
• Science Fiction
• Historical & Documentary
• Crime & Mystery

Language Register

• Primarily conversational German (informal to neutral)
• Diverse dialects and regional variations
• Emotional expressions and colloquialisms
• Natural sentence fragments (as spoken in dialogue)

Limitations

• Context missing: Single subtitle lines without surrounding context
• Encoding artifacts: Some lines may contain special characters (Ã, etc.)
• Duplicates: 7.5% duplicates (mostly natural dialogue repetitions)
• Genre bias: Over-representation of popular film/TV genres
• Translation artifacts: Some subtitles are translations, not native German

Ethical Considerations

• Subtitles may contain offensive language, violence references, or mature content
• No personally identifiable information (PII) should be present
• Content reflects diverse social contexts from movies/TV shows

License

The licensing status of this dataset is unknown. OPUS OpenSubtitles data is derived from publicly available subtitle files. Users should verify compliance with applicable copyright and licensing requirements before use.

Citation

If you use this dataset, please cite:

@dataset{german_opus_subtitles_2025,
  title={German OPUS OpenSubtitles},
  author={arnomatic},
  year={2025},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/arnomatic/german-opus-subtitles}
}

Acknowledgments

• OPUS Project for collecting and distributing subtitle data
• OpenSubtitles.org for the original subtitle files
• All contributors to the OPUS corpus

Contact

For questions or issues, please open an issue on the dataset repository.

---

Dataset Version: 1.0 Last Updated: 2025-01-12 Maintainer: arnomatic