Skip to content

Dataset · huuuyeah

City council meeting transcripts and minutes benchmark

huuuyeah/meetingbank

Long transcripts, video, and minutes from 1,366 council meetings for summarization evaluation.

What's inside

An English-language benchmark compiled from recordings of city council meetings across 6 major US cities. It contains 1,366 meetings with a total duration of over 3,579 hours of video, along with text transcripts, agendas, and PDF documents containing minutes. The average meeting lasts 2.6 hours, and its transcript volume exceeds 28,000 tokens.

What you can do with it

  • Evaluating and training models for summarizing ultra-long texts
  • Training systems to extract structured summaries and agendas from transcripts
Tasks
Summarization, Text generation
Languages
English
Size
1K–10K rows
Downloads this month
1.3K
Likes
38

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("huuuyeah/meetingbank")