CSI
Introduced by Mael et al. in Graph2Speak: Improving Speaker Identification using Network Knowledge in Criminal Conversational Data
CSI is a criminal conversational dataset for speaker identification built from the CSI television show. The authors collected transcripts of 39 episodes and video/audio of 4 episodes. Each episode involves on average more than 30 speakers. Utterances last on average 3 to 4 seconds. There are around 45 to 50 distinct scenes/conversations per episode.
Papers
| Paper | Code | Results | Date | Stars |
|---|
