Ru Support Toxicity Detection: Balanced Russian Text for Binary Classification

A balanced dataset constructed from five distinct Russian-language sources for binary toxicity classification. The author Nelera compiled toxic comments from 'Russian Toxic Comments' and neutral content from 'MTSBerquad LFQA'. The dataset page was last updated on July 17, 2026.

Use Cases

• Train binary classifiers to detect toxic Russian text based on the described balanced sample.
• Benchmark toxicity detection models against a dataset sourced from multiple Russian-language platforms.
• Fine-tune language models for Russian content safety using the explicitly constructed classification data.

Strengths

• Data is explicitly described as a balanced sample, which can help mitigate class imbalance issues.
• Sourced from five distinct Russian-language sources, suggesting potential diversity in text style and origin.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale training.
• Freshness should be verified as the last update timestamp is in the future (2026-07-17).

Provenance

Source

Aggregated from Russian Toxic Comments (klamas/russian-toxic) and MTSBerquad LFQA (MTS-AI-SearchSkill/MTSBerquad), plus three other unspecified Russian-language sources.

Collection Method

Constructed as a balanced sample from multiple sources.

Freshness

Last updated 2026-07-17 18:56:04

Geography

Russian-language content; specific geographic coverage is unknown.