A two-speaker dataset manually aligned at the phoneme level, providing ground truth for phonetic alignment research. It contains 200 instances from both a male and a female speaker. The dataset was created by falabrasil and has been used as ground truth in multiple academic papers from 2016 to 2022.
Use Cases
- Training forced phonetic alignment models based on manually aligned phoneme-level ground truth.
- Evaluating the performance of speech alignment algorithms on both male and female speakers.
- Studying speaker-specific phonetic variations using the provided male and female speaker data.
Strengths
- Contains 200 phoneme-aligned instances from each of a male and a female speaker.
- Manually aligned at the phoneme level, providing high-quality ground truth.
- Has been used as ground truth in at least five published academic papers.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Row count is unknown, which may limit suitability assessment.
- Column-level documentation is absent; field semantics must be inferred after download.
Provenance
- Source
- falabrasil on Hugging Face
- Collection Method
- Manually aligned at the phoneme level.
- Freshness
- Last updated 2026-06-15 13:24:52; freshness should be verified.