Kinopotok Toxicity: Russian Customer Support Messages for Toxic Text Classification

Kinopotok Toxicity is a training prototype dataset for filtering toxic messages in the support chat of the online cinema 'Kinopotok'. The dataset contains 120 records for binary classification of text as toxic or non-toxic. It was created by author petaevd and was last updated on 2026-07-15.

Use Cases

• Train a binary text classifier based on the described toxic/non_toxic labels.
• Experiment with moderation models for customer support chats based on the described use case.
• Benchmark model performance on a small, focused dataset for educational purposes based on the 120-record size.
• Study patterns of insults, threats, and profanity in Russian-language online communications based on the description.

Strengths

• Contains 120 records for training.
• Designed for a specific, practical task: filtering toxic messages in a customer support chat.
• Includes a clear binary classification target (toxic/non_toxic).

Limitations

• The dataset is small with only 120 records, limiting its use for training complex models.
• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface, author petaevd

Collection Method

Likely collected from or simulated for the support chat of the online cinema 'Kinopotok'.

Freshness

Last updated 2026-07-15 09:04:21; freshness should be verified.