Observed arrival · 2026-11-05
Marani’s open ledger for Georgian speech
Marani documents the collection and processing of Georgian speech data, alongside evaluation results for recognition models.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Georgian speech researchers and speech-model builders
- Worth noticing
- The source table separates audio cleared for public release from material marked “no” or “not cleared.”
Field notes
The pipeline separates collection, cleaning, transcription, dataset construction, and model evaluation, with progress counts attached to each stage. Its benchmark notes specify a 5.5-hour Parliament test held out from training and a separate FLEURS test; the page also reports compute costs and job activity. Human transcript review is marked as not started, so the displayed model results are not presented as a fully human-reviewed dataset.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue