How mahtafetrat/gptinformal-persian-speech-dataset works
GPTInformal-Persian is a speech dataset: GPT-4o-generated informal Persian text is recorded as speech, then audio and text are force-aligned into smaller pairs published as dataset data. The README describes the resulting Parquet columns and intended speech-related uses, and points users to Hugging Face for download. Only the README is available here, so these nodes represent documented stages and capabilities, not inspected implementation code; relationships shown are limited to those it explicitly describes.
Main entry points: README.md Informal text
Components 8
Data creation
- Informal text
README.md - Speech recording
README.md - aeneas alignment
README.md
Dataset and access
- Audio-text chunks
README.md - Parquet dataset
README.md
Use
- Speech tasks
README.md
Other components
- GPT-4o
- Hugging Face
Connections 8
- GPT-4o → Informal text: generates
README.md - Informal text → Speech recording: recorded as
README.md - Speech recording → aeneas alignment: force-aligned with
README.md - Informal text → aeneas alignment: force-aligned with
README.md - aeneas alignment → Audio-text chunks: produces
README.md - Audio-text chunks → Parquet dataset: presented in
README.md - Parquet dataset → Hugging Face: download from
README.md - Parquet dataset → Speech tasks: supports
README.md