Skip to content

Dataset · winvoker

Three-class Turkish sentiment analysis dataset

winvoker/turkish-sentiment-analysis-dataset

Corpus of 100 thousand to 1 million Turkish sentences labeled as positive, negative, and neutral.

What's inside

Contains between 100 thousand and 1 million Turkish texts for a three-class sentiment classification task. Unlike standard binary sets, a neutral class has been added here, with examples extracted from a Turkish Wikipedia dump and supplemented with random texts and data from several open sources.

What you can do with it

  • Training Turkish sentiment classifiers taking neutral phrases into account
  • Fine-tuning multilingual language models for review and social media analysis
  • Filtering emotionally neutral content in user message streams
Tasks
Text classification
Languages
Turkish
Size
100K–1M rows
Downloads this month
1.2K
Likes
50

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("winvoker/turkish-sentiment-analysis-dataset")