pttrn-io/ubuntu-irc-days

Ubuntu IRC channel-days

Five years of two Ubuntu IRC channels, one row per channel-day, used as the showcase corpus for the MADS Data Analysis and Visualisation course.

What one row is

One row is one channel on one day. The date is a column; the individual message times sit inside the text, one message per line:

column	type	meaning
created	datetime64[ms]	the day
channel	string	#ubuntu-uk or #ubuntu-nl
text	string	every message that day, [HH:MM] <nick> message per line

Turning that into (timestamp, author, message) rows is deliberately not done here — it is the first exercise of the course, and shipping it pre-parsed would remove the lesson. Expect to write a regex.

The slice

channel	days	messages	per day	peak/trough hour
#ubuntu-nl	1,545	116,342	75	141.5×
#ubuntu-uk	1,826	510,834	280	15.3×

Selected from common-pile/ubuntu_irc by scripts/build_irc_showcase.py in the course repository, taking the two channels above between 2013-01-01 and 2017-12-31.

#ubuntu-uk is the primary channel: English, busy enough for daily and weekly structure to be visible, with a release-day spike on every Ubuntu release in the window. #ubuntu-nl is deliberately thin — it is what a quiet channel looks like, and what small sample sizes cost you.

The global #ubuntu channel is the control for the daily-cycle contrast: it runs 24/7 across every timezone, so its peak-to-trough ratio is close to flat where a regional channel's is dramatic. Its raw text is ~400 MB, so only an hourly aggregate is distributed, in the course repository rather than here.

Loading it

import pandas as pd

days = pd.read_parquet(
    "hf://datasets/pttrn-io/ubuntu-irc-days/ubuntu_irc_days.parquet"
)

Provenance and licence

Derived from common-pile/ubuntu_irc, which is Public Domain; this slice is released under CC0-1.0 to match. The messages are public IRC logs from support channels, published by the Ubuntu project. They are real people's words: quote them as you would any published archive, and do not use them to profile individuals.