sronquillo2/Enhanced_Emotion_Classification_Dataset

1---2language:3- en4pretty_name: Enhanced Emotion Classification Dataset5version: 2.06tags:7- text-classification8- emotion9- sentiment-analysis10- ekman-emotions11- text12license: mit13task_categories:14- text-classification15task_ids:16- sentiment-classification17---18 19# Enhanced Emotion Classification Dataset (v2.0)20 21## Dataset Description22 23This dataset is an enhanced version of the emotion classification dataset, including multiple sources of emotion data with Ekman emotion mapping. It contains a total of 240,426 samples across 7 emotion categories, with each sample labeled with its original data source.24 25### Dataset Structure26 27The dataset is split into three parts:28- **Train**: 186,619 samples (77.6%)29- **Validation**: 31,086 samples (12.9%)30- **Test**: 22,721 samples (9.4%)31 32### Emotion Categories33 34The dataset includes 7 Ekman basic emotions:35- neutral36- joy37- sadness38- anger39- fear40- surprise41- disgust42 43## Data Sources44 45The dataset combines data from 7 different sources:46 471. **Movies_Reviews**48   - Movie review emotion data49   - Contains 7 emotion categories50 512. **DailyDialog**52   - Real dialog data with multi-turn conversations53   - Contains 7 emotion categories54 553. **GoEmotions**56   - Reddit comment data with colloquial expressions57   - Contains 7 emotion categories58 594. **ISEAR**60   - International emotion research data with high-quality text61   - Contains 7 emotion categories62 635. **MELD**64   - Multimodal emotion dialog data from the TV show "Friends"65   - Contains 7 emotion categories66 676. **mteb_emotion**68   - Emotion analysis dataset with various emotion expressions69   - Contains 7 emotion categories70 717. **Tweet Emotions**72   - Twitter tweet data including @mentions73   - Contains 7 emotion categories74 75## Data Format76 77Each CSV file contains the following columns:78- `text`: Text content79- `label_text`: Emotion label (one of the 7 emotion categories)80- `source`: Data source identifier (e.g., dailydialog, goemotions, tweetemotions, etc.)81 82## File Structure83 84```85dataset_huggingface_enhance/86├── train.csv              # Merged training set (186,619 samples)87├── val.csv                # Merged validation set (31,086 samples)88├── test.csv               # Merged test set (22,721 samples)89├── README.md              # Dataset documentation90├── dataset_infos.json     # Hugging Face dataset configuration91├── info.md                # Detailed dataset statistics92├── label_list.txt         # List of emotion labels93└── stats.py               # Dataset statistics script94```95 96## Usage97 98To use this dataset with Hugging Face Datasets library:99 100```python101from datasets import load_dataset102 103# Load the dataset104dataset = load_dataset("jiangchengchengNLP/Enhanced_Emotion_Classification_Dataset")105 106# Access specific splits107train_dataset = dataset["train"]108val_dataset = dataset["validation"]109test_dataset = dataset["test"]110```111 112## Emotion Distribution113 114### Training Set (186,619 samples)115- neutral: 51.11%116- joy: 21.64%117- sadness: 7.98%118- anger: 5.97%119- fear: 5.90%120- surprise: 5.11%121- disgust: 2.29%122 123### Validation Set (31,086 samples)124- neutral: 40.96%125- joy: 24.15%126- sadness: 9.38%127- fear: 8.06%128- anger: 7.49%129- surprise: 7.06%130- disgust: 2.89%131 132### Test Set (22,721 samples)133- neutral: 45.11%134- joy: 22.88%135- sadness: 9.13%136- anger: 7.61%137- fear: 6.67%138- surprise: 5.93%139- disgust: 2.67%140 141## Notes142 143- The dataset has class imbalance, with neutral being the most common category (~40-51%) and disgust being the least common (~2-3%).144- Each sample includes a `source` field indicating its original data source, which allows for source-specific analysis.145- Different data sources have different text styles, which may affect model performance.146- The dataset uses Ekman emotion mapping, which maps various emotion labels to the 7 basic emotions.147 148## License149 150The dataset is released under the MIT License.151 152## Citation153 154If you use this dataset in your research, please cite the original datasets:155 156- DailyDialog: https://aclanthology.org/I17-1099/157- GoEmotions: https://arxiv.org/abs/2005.00547158- Tweet Emotions: https://www.aclweb.org/anthology/W18-6212/159- MELD: https://arxiv.org/abs/1810.02508160- ISEAR: https://link.springer.com/article/10.1007/BF02112196161 162## Version History163 164- **v2.0** (2026-01-07): Updated dataset with merged sources and source field165  - Total samples: 240,426166  - Added `source` field to all samples167  - Updated emotion distribution statistics168  - Improved data quality and consistency169  - Filtered all NaN values from the dataset170 171- **v1.0**: Initial release172 173