datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
sst2
Dataset Card for [Dataset Name]
Dataset Summary
The Stanford Sentiment Treebank is a corpus with fully labeled parse trees that allows for a complete analysis of the
compositional effects of sentiment in language. The corpus is based on the dataset introduced by Pang and Lee (2005)
and consists of 11,855 single sentences extracted from movie reviews. It was parsed with the Stanford parser and
includes a total of 215,154 unique phrases from those parse trees, each… See the full description on the dataset page: https://huggingface.co/datasets/stanfordnlp/sst2.sst5
Stanford Sentiment Treebank - Fine-Grained
Stanford Sentiment Treebank with 5 labels: very positive, positive, neutral, negative, very negative
Splits are from:
https://github.com/AcademiaSinicaNLPLab/sentiment_dataset/tree/master/data
Training data is on sentence level, not on phrase level!
sst2
Stanford Sentiment Treebank - Binary
Stanford Sentiment Treebank with 2 labels: negative, positive
Splits are from:
https://github.com/AcademiaSinicaNLPLab/sentiment_dataset/tree/master/data
Training data is on sentence level, not on phrase level!
sst2sst2_ptThe Stanford Sentiment Treebank consists of sentences from movie reviews and
human annotations of their sentiment. The task is to predict the sentiment of a
given sentence. We use the two-way (positive/negative) class split, and use only
sentence-level labels.common_voice_enSST2Perturbed
Dataset Card for "SST2Perturbed"
More Information needed
sstirsst1Dataset used in the paper:
A thorough benchmark of automatic text classification
From traditional approaches to large language models
https://github.com/waashk/atcBench
To guarantee the reproducibility of the obtained results, the dataset and its respective CV train-test partitions is available here.
Each dataset contains the following files:
data.parquet: pandas DataFrame with texts and associated encoded labels for each document.
split_<k>.pkl: pandas DataFrame with k-cross validation… See the full description on the dataset page: https://huggingface.co/datasets/waashk/sst1.sst5-paraphrased
sst5-paraphrased
Paraphrased version of the Stanford Sentiment Treebank v5 (SST-5) dataset. Each sentence has been paraphrased while preserving its original fine-grained sentiment label.
Model
Paraphrases were generated using Qwen/Qwen3-30B-A3B-Instruct-2507.
Original Dataset
Source: SetFit/sst5
Task: Fine-grained Sentiment Classification
Classes: 5 (0 = very negative, 1 = negative, 2 = neutral, 3 = positive, 4 = very positive)
Dataset Structure… See the full description on the dataset page: https://huggingface.co/datasets/concretejungles/sst5-paraphrased.task363_sst2_polarity_classification
Dataset Card for Natural Instructions (https://github.com/allenai/natural-instructions) Task: task363_sst2_polarity_classification
Additional Information
Citation Information
The following paper introduces the corpus in detail. If you use the corpus in published work, please cite it:
@misc{wang2022supernaturalinstructionsgeneralizationdeclarativeinstructions,
title={Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks}… See the full description on the dataset page: https://huggingface.co/datasets/Lots-of-LoRAs/task363_sst2_polarity_classification.fewshot-prompted-sst2augmented-glue-sst2
Dataset Card for Augmented-GLUE-SST2
Automatically augmented data from train split of SST-2 dataset using conditional text generation approach.
Code used to generate this file will be soon available at https://github.com/IntelLabs/nlp-architect.
SST2_train67k_test1.8k_valid0.8k
Dataset Card for "SST2_train67k_test1.8k_valid0.8k"
More Information needed
sst2_with_predictions
Comparing model predictions and ground truth labels with Rubrix and Hugging Face
Build dataset
You can skip this step if you run:
from datasets import load_dataset
import rubrix as rb
ds = rb.DatasetForTextClassification.from_datasets(load_dataset("rubrix/sst2_with_predictions", split="train"))
Otherwise, the following cell will run the pipeline over the training set and store labels and predictions.
from datasets import load_dataset
from transformers import pipeline… See the full description on the dataset page: https://huggingface.co/datasets/rubrix/sst2_with_predictions.SSTQAsst2-textbugger
Stanford Sentiment Treebank - Binary
sst2-pwws
Stanford Sentiment Treebank - Binary
sst2sStoryClozeThis dataset was introduced in https://github.com/slp-rl/SpokenStoryCloze.
@misc{hassid2024textuallypretrainedspeechlanguage,
title={Textually Pretrained Speech Language Models},
author={Michael Hassid and Tal Remez and Tu Anh Nguyen and Itai Gat and Alexis Conneau and Felix Kreuk and Jade Copet and Alexandre Defossez and Gabriel Synnaeve and Emmanuel Dupoux and Roy Schwartz and Yossi Adi},
year={2024},
eprint={2305.13009},
archivePrefix={arXiv}… See the full description on the dataset page: https://huggingface.co/datasets/speed/sStoryCloze.sst2-openaiOriginal source: https://github.com/openai/generating-reviews-discovering-sentiment
This dataset is different from the dataset distributed by GLUE, which means the metric shouldn't be compared with the SST2 performance in GLUE.
The description of SST2 dataset in the paper is the following.
The Stanford Sentiment Treebank (SST)(Socher et al., 2013) was created specifically to evaluate more complex compositional models of language. It is de-rived from the same base dataset as MR but was… See the full description on the dataset page: https://huggingface.co/datasets/sh0416/sst2-openai.sst2-textfooler
Stanford Sentiment Treebank - Binary
sst2-audiosst5
Stanford Sentiment Treebank - Fine-Grained
Stanford Sentiment Treebank with 5 labels: very positive, positive, neutral, negative, very negative
Splits are from:
https://github.com/AcademiaSinicaNLPLab/sentiment_dataset/tree/master/data
Training data is on sentence level, not on phrase level!
glue-sst2
Dataset Card for "glue-sst2"
More Information needed
Note: This dataset was utilized for the evaluation of probability-based prompt selection techniques in the paper 'Improving Probability-based Prompt Selection Through Unified Evaluation and Analysis'. It differs from the actual benchmark dataset.
SST-2-attrpromptThis is the data used in the paper Large Language Model as Attributed Training Data Generator: A Tale of Diversity and Bias.
label.txt: the label name for each class
train.jsonl: The original training set.
valid.jsonl: The original validation set.
test.jsonl: The original test set.
simprompt.jsonl: The training data generated by the simple prompt.
attrprompt.jsonl: The training data generated by the attributed prompt.
sst2_corpusverl-distill-datasetglue_augmented_sst2
Dataset Card for glue_augmented_sst2
Dataset Description
Augmented SST-2 dataset
Reference: https://huggingface.co/datasets/glue
kyrgyz-sst2
Kyrgyz SST-2
Task
Sentiment Classification (Binary Sentence Classification)
Description
Stanford Sentiment Treebank binary classification task translated from English to Kyrgyz. Each entry contains an English sentence, its Kyrgyz translation, and a binary sentiment label.
Labels: negative, positive
Format: JSONL with fields such as sentence, sentence_ky, and label
Dataset Size
Split
Entries
Train
6,920
Validation
872
Test
1,821… See the full description on the dataset page: https://huggingface.co/datasets/metinovadilet/kyrgyz-sst2.