| # Dataset Card: Multi-Task German Text Classification Dataset | |
| ## Dataset Description | |
| This dataset contains German text samples labeled for three tasks: | |
| 1. **Fake News Detection (`is_fake`)** | |
| 2. **Hate Speech Detection (`is_hate_speech`)** | |
| 3. **Toxicity Detection (`is_toxic`)** | |
| Each entry in the dataset has binary or missing labels for the respective tasks: | |
| - `0`: Negative | |
| - `1`: Positive | |
| - `-1`: Not labeled for the task | |
| The dataset is useful for training and evaluating models on multi-task learning objectives in the context of online German text. | |
| --- | |
| ## Dataset Structure | |
| | **Column** | **Description** | | |
| |-----------------------|-----------------------------------------------------------| | |
| | `text` | The German text sample | | |
| | `is_fake` | Binary label for fake news detection (0, 1 or -1) | | |
| | `is_hate_speech` | Binary label for hate speech detection (0, 1, or -1) | | |
| | `is_toxic` | Binary label for toxicity detection (0, 1, or -1) | | |
| ### Sample Data | |
| This is only top few item of dataset where fake news is given as 0 or 1 but other 2 which are going to be hidden are in -1. similarly for hatespeech and toxicity , other 2 are -1 | |
| | **Text** | `is_fake` | `is_hate_speech` | `is_toxic` | | |
| |-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------|------------------|------------| | |
| | Im Studio von @1730Sat1live sprechen wir heute mit Stefan Schulte, dem Vorstandsvorsitzenden der @FraportAG, über die Lage am Frankfurter Flughafen... | 0 | -1 | -1 | | |
| | gigi hat einfach überhaupt keine lust mehr auf lucas. gigi 🤝 wir alle #ibes | 0 | -1 | -1 | | |
| | In Österreich 🇦🇹 ist Bundespräsident @vanderbellen für eine 2. Amtszeit angelobt worden... | 0 | -1 | -1 | | |
| | Was ist der Mindestlohn pro Stunde in Vietnam? https://t.co/VQof8kwRlL | 0 | -1 | -1 | | |
| | #GretaThunberg will jetzt den #Kapitalismus überwinden... | 0 | -1 | -1 | | |
| --- | |
| ## Usage | |
| This dataset is designed for training and evaluating multi-task classification models in German. The tasks include: | |
| - Fake news detection | |
| - Hate speech identification | |
| - Toxicity assessment | |
| ### Example Usage in Python | |
| ```python | |
| import pandas as pd | |
| # Load the dataset | |
| df = pd.read_csv("path_to_dataset.csv") | |
| # Access text and labels | |
| for index, row in df.iterrows(): | |
| text = row['text'] | |
| is_fake = row['is_fake'] | |
| is_hate_speech = row['is_hate_speech'] | |
| is_toxic = row['is_toxic'] | |
| # Example: Print the data | |
| print(f"Text: {text}, Fake: {is_fake}, Hate Speech: {is_hate_speech}, Toxic: {is_toxic}") | |
| ``` | |