- Added: age.[train|dev|test] tables with speaker information
- consists of randomly selected 20 emotionally neutral samples per speaker
- all tables being age/gender balanced
- Added: Tables with vote frequencies
- Added: Gold standard tables including
no_agreement
samples - Added:
files
table to replacespeaker
table with corrupted indicator column - Changed: Rename emotion tables to default emotion table names
- Changed: Rename scheme
emotion.value
toemotion.level
- Changed: Add index
0
to first winning emotion columns - Removed: Corrupted file with no audio is removed from index in emotion tables
- Removed:
speaker
table - Removed:
emotion.agreement
column inemotion.categories.<modality>.<split>
- Changed: set usage to
commercial
- Added: Unofficial speaker-independent train, dev, test splits for each rating modality and the desired emotion
- Removed: Tables
emotion
,emotion.face
,emotion.voice
,emotion.face
- Added: author and license information
- Added: conversion to new audb format
- Fixed: speaker table
- Fixed: speaker scheme
- Fixed: maven location
- Initial release