Datasets:
metadata
annotations_creators: []
language:
- en
language_creators: []
license: []
multilinguality:
- monolingual
pretty_name: adaption-emotion
size_categories:
- 10K<n<100K
source_datasets:
- extended|https://huggingface.co/datasets/dair-ai/emotion
tags:
- adaption
- instruction-tuning
- medical
- personal-growth
- fashion-beauty
task_categories: []
task_ids: []
This dataset is a remastered version of this dataset prepared using Adaption's Adaptive Data platform.
adaption-emotion
This dataset contains English text samples labeled with one of six basic emotions: sadness, joy, love, anger, fear, and surprise. Designed for multi-class text classification tasks, it features over 16,000 training examples derived from social media or blog posts expressing personal feelings. The data is split into training, validation, and test sets to facilitate model development and evaluation.
Dataset size
There are 17,000 data points in this dataset. This is an instruction tuning dataset.
Quality of Remastered Dataset
The final quality is A, with a relative quality improvement of 350.0%.
Domain
- Medical (40%)
- Personal-growth (24%)
- Fashion-beauty (6%)
Language
- English (100%)
Tone
- Empathetic (30%)
- Helpful (22%)
- Empowering (12%)
Evaluation Results
Quality Gains:
Grade Improvement:
Percentile Chart:
