Podcast annotation dataset for paper "Identifying Introductions in Podcast Episodes from Automatically Generated Transcripts "

Dataset for paper "Identifying Introductions in Podcast Episodes from Automatically Generated Transcripts". Please refer to the paper for details. Compared to the dataset used in the paper, 20 out of the 445 episodes have been removed due to copyright issues. 

Each data file contains the following fields:

- "episode_intro_start": the time stamp for episode introduction start (in milliseconds)

- "episode_intro_end": the time stamp for episode introduction end (in milliseconds)

- "program_intro_start": the time stamp for program introduction start (in milliseconds)

- "program_intro_end": the time stamp for program introduction end (in milliseconds)

- "program_name": name of the podcast program

- "episode_name": name of the podcast episode

- "transcription": JSON string containing the transcription, including the timestamps.

- "annotator": anonymized annotator ID.