Ru Toxicity Support Binary: Russian-Language Toxic Message Detection for Support Chats

A Russian-language dataset for binary classification of toxic messages in support chat conversations. The dataset was created by Izbebe and last updated on July 16, 2026. It contains a stratified 80/10/10 train/validation/test split with balanced classes.

Use Cases

• Train binary classifiers for toxic message detection based on Russian text from support chats.
• Benchmark model performance on stratified and balanced splits for toxicity classification.
• Fine-tune pre-trained language models for Russian-language moderation tasks using the provided text and label fields.

Strengths

• Dataset is stratified and class-balanced across its train, validation, and test splits.
• Sources for the toxic class (label 1) are explicitly named: textdetox/multilingual_toxicity_dataset (ru) and the toxic part of s-nlp/ru_paradetox.

Limitations

• Row count, file formats, and column-level documentation are unknown, which limits suitability assessment.
• Description metadata is limited; actual data quality and the source for the non-toxic class require manual inspection after download.

Provenance

Source

huggingface, author Izbebe

Collection Method

Combined from existing datasets: textdetox/multilingual_toxicity_dataset (ru) and s-nlp/ru_paradetox for toxic examples; source for non-toxic examples is not fully specified.

Freshness

Last updated 2026-07-16 21:32:39; freshness should be verified.