Russian Emotional Speech Dialogues from Actor Performances

Approximately 3.5 hours of Russian speech recordings where actors performed dialogues expressing pre-distributed emotions. Each actor voiced an emotion for about 3 minutes. The dataset is intended for audio emotion recognition tasks.

Use Cases

• Train an audio classification model for emotion recognition using the Russian speech recordings.
• Analyze acoustic features from the actor-performed emotional speech to identify patterns for specific emotional states.
• Benchmark emotion recognition algorithms on a monolingual Russian dataset of expert-generated emotional dialogues.

Strengths

• Contains approximately 3.5 hours of live speech data.
• Emotional labels are expert-generated by actors performing pre-distributed emotions.
• Data is provided in Parquet format, suitable for use with libraries like Polars and Pandas.

Limitations

• The dataset size is categorized as 1K<n<10K, indicating a relatively small sample size for training complex models.
• Specific data fields, instance counts, and detailed structure are not provided in the input.
• The dataset is monolingual (Russian), limiting applicability to other languages.

Provenance

Source

Hugging Face dataset authored by Aniemore.

Collection Method

Assembled from live speech recordings by actors voicing pre-distributed emotions in dialogue.

Freshness

Last updated on 2023-06-10.

Geography

Region is indicated as 'us', but the language is Russian. Specific geographic coverage is unclear.